On Fri, Jul 13, 2007 at 09:35:09 +0100, Malcolm Wallace wrote:

dons@cse.unsw.edu.au (Donald Bruce Stewart) writes:

...

As we sit here riding the Haskell wave:

http://www.cse.unsw.edu.au/~dons/tmp/cafe.png

with nearly 2000 (!) people reading haskell-cafe@, perhaps its time to think some more about how to build and maintain this lovely Haskell community we have.

Yes, the sheer volume of posts is definitely becoming a problem (for me, at least). All your suggestions for keeping the community polite and helpful are good. But I wonder if there are also any useful technical tips for users like myself, who would like to be able to keep up, but feel they are gradually drowning?

One obvious solution is to split the list into several, more specialised lists. It's far from obvious, at least to me, how to do that with this list though. Personally I try to read the first post in every conversation thread. If it doesn't grab be then I delete the whole thread. Not an ideal strategy, but it helps me keeping my day job ;-) /M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus@therning.org Jabber: magnus.therning@gmail.com http://therning.org/magnus

Lutz Donnerhacke wrote:

Switch to Usenet. The new haskell group will die, if the traffic will not increase.

If anything, Usenet is even worse than mailing lists for volume, especially of spam. Also, very few sites maintain their nntp servers adequately these days - e.g. comp.lang.haskell has never made it to where I work. Regards, Malcolm

Malcolm Wallace said that Usenet is pretty bad these days. Lutz Donnerhacke disagreed. Living as near the Antarctic as I do, I find that the Internet works fine, e-mail is just great, but NNTP is effectively dead. I stopped reading net news (with sadness and reluctance) about eight years ago when I found that I couldn't rely on getting comp.arch any more. It was also sad to say goodbye to comp.lang.functional, but that hadn't been heard from for some time either. I was very happy to discover Haskell-Café and would be sorry to lose it, as I would if it went over to NNTP.

Chris Smith wrote:

Well, it doesn't have to "go over" anywhere. I'm reading and posting just fine with NNTP right now. It works great.

How'd you manage that?

I found out I could do so by reading this thread. Until then, I'd avoided haskell-cafe, hanging out mostly on IRC for the last few months because I didn't want the high volume of email.

I've been avoiding it for over a year for the same reason. (And because I didn't want people to have my real email address, but never mind...) Actually, since this is the first time I've tried to seriously use Thundrebird for email, I'm surprised at how buggy it is...

Malcolm Wallace writes:

Yes, the sheer volume of posts is definitely becoming a problem (for me, at least).

As a newcomer I was stunned that this otherwise very sophisticated community was using an email list rather than a bulletin board. The shear torrent of email was impacting my mail program performance. Then I chased somes repies leading to earlier threads, and found bulletin-board stye access to our list at http://thread.gmane.org/gmane.comp.lang.haskell.cafe/ I much prefer this access, even if I wish the window spit vertically rather than horizontally. A bulletin board has the capabibility to evolve, e.g. into multiple entry points. Are there statistics on who reads which way? Posting stats should be obvious with a little script, for anyone who still has a hoard of messagess at homne.

Jules Bean wrote:

I find it incredibly surprising whenever I discover than an otherwise sophisticated community has adopted a bulletin board rather than email ;)

Erm... why?

Conversely, a bulletin board cannot be read offline, traps users into a single UI (in every case I've used, slow, ugly, and lacking in functionality) over which they have no control.

Oh... I suppose. Still, a newsreader doesn't appear to have those limitations.

Modern email programs have sophisticated sorting, filtering, scoring, processes and they allow me to read messages while offline, search them locally, etc etc. They have customisable key bindings, they allow me to read all of my mailing lists in one place, they are scriptable, may support plugins... all of this under full user control.

I'd just prefer not to have to wait through 100 emails a day to find the few that interest me. With a newsreader, I can simply mark "ignore" on the threads that don't interest me, and I'm done. (Also, either my ISP or Thunderbird itself keeps magically "loosing" certain emails, which is rather irritating...)

On Fri, 13 Jul 2007, Dave Bayer wrote:

As a newcomer I was stunned that this otherwise very sophisticated community was using an email list rather than a bulletin board. The shear torrent of email was impacting my mail program performance.

This is a cultural thing, and assuming that it's a lack of sophistication on our part is a bad idea - on the contrary, some of the better reasons to avoid a web-based board are entirely about enabling sophistication. Boards need polling. Boards force a single user interface on everyone. Boards don't enable local archives. Etc etc etc. With mail, you can pick a client to suit your needs.

A bulletin board has the capabibility to evolve, e.g. into multiple entry points.

So does a mailing list when you own the domain it's hosted on. -- flippa@flippac.org Sometimes you gotta fight fire with fire. Most of the time you just get burnt worse though.

On 7/13/07, Philippa Cowderoy wrote:

This is a cultural thing, and assuming that it's a lack of sophistication on our part is a bad idea - on the contrary, some of the better reasons to avoid a web-based board are entirely about enabling sophistication.

There's a very simple argument for using a bulletin-board: bulletin boards

are the lazy solution.

Andrew Coppin wrote:

Dave Bayer wrote:

...

Malcolm Wallace writes:

...

Yes, the sheer volume of posts is definitely becoming a problem (for me, at least).

As a newcomer I was stunned that this otherwise very sophisticated community was using an email list rather than a bulletin board. The shear torrent of email was impacting my mail program performance.

Seconded. (I also commented on it at the time, and was told "use IRC or SMTP". Personally I'd prefer NNTP, but apparently I'm a minority here...)

What are you on about? Use gmane.org... -- Bardur Arantsson Nuke unborn gay whales for Jebus! "Most Broadly Offensive Slogan Ever"

On Fri, Jul 13, 2007 at 09:35:09AM +0100, Malcolm Wallace wrote:

Yes, the sheer volume of posts is definitely becoming a problem (for me, at least).

The Haskell lists are quite peculiarly named; the haskell@ list is pretty much what would be haskell-announce@ anywhere else, and haskell-cafe@ what would be haskell@ elsewhere. A haskell-cafe@ list elsewhere would probably be where the haskell@ people discuss things which aren't actually related to Haskell (like e.g. the demon.local newsgroup). I think the number of posts in the wrong place would be lower if these were more conventionally named (although there aren't a lot of them anyway). I think it would make sense to: * Rename haskell@ to haskell-announce@, and redirect mails from haskell@ to haskell-announce@ for some period. * After that, rename haskell-cafe@ to haskell@ in a similar manner. * At any point, create haskell-new-users@. This would have the advantage that people might not be so intimidated at making their first post here, and posts wouldn't be answered with category theory or scary type extensions. The disadvantages are that it makes an artificial barrier (when is someone ready to post to haskell@ instead?) and it doesn't split the discussion up by topic (if you want to see all the discussion about X, you need to follow both lists so you have the same amount of mail to read). I'm not sure if this is a good idea. * At any point, create topic-specific lists, e.g. a list for discussion type system extensions (not sure if that's a good one, as often an extension is the /answer/ to a question "how do I do foo?"). This needs people to work out which lists would allow us to make a significant dent to the haskell-cafe traffic. Thanks Ian, also drowning

Hi

...

* At any point, create haskell-new-users@. This would have the advantage that people might not be so intimidated at making their first post here, and posts wouldn't be answered with category theory or scary type extensions. The disadvantages are that it makes an artificial barrier (when is someone ready to post to haskell@ instead?)

Just an idea: How about haskell-first-post@ to encourage the second post to go to haskell@? Or is that too dictator-like?

Why not learning@haskell.org, for people who aren't using Haskell to do things, but are working through tutorials. A number of questions relate directly to books such as SOE, and these could be easily answered there. Thanks Neil

Ian Lynagh wrote:

I think the number of posts in the wrong place would be lower if these were more conventionally named (although there aren't a lot of them anyway).

There are very few inappropriate posts to the haskell@ list. I very much doubt that the list names are a problem.

* At any point, create haskell-new-users@.

This would pretty-much guarantee that no-one with the necessary expertise to answer questions was actually listening to them. :-( Regards, Malcolm

Jim Burton wrote:

Very timely! It's sad that haskell-cafe has so much noise now.

I disagree with that characterisation. I don't mean to be pedantic, but I don't think haskell-cafe has lots of noise. I think it has lots of signal! Quite different. We don't have a problem (in my perception, at least) with the quality of either posts of responses. Not a general one, anyway. The only problem is that the volume is increasing; which is a problem if that makes it hard for valuable contributors to keep contributing.

As well as being nice, can't you sometimes tell people to RTFM?

The problems people have with haskell are often conceptual, and the manual doesn't help them, because they don't (yet) understand the language well enough to understand the manual. I very, very rarely see a question here about 'how to use a library funciton' or similar which could, in fact, be easily looked up in a manual. People do quite often respond to posts with links to the online library documentation, which is great.

Or, You've asked that before,

That's certainly a fair thing to say, if it's true. I don't see that happening very often.

One way of protecting the community is to protect this list from drowning in noise and being a bit rough with newbies who don't do any research at all before asking is perfectly acceptable in my view.

I disagree with that on two separate levels: (a) I don't think being rough with newbies is the right response. (b) I also don't think it would achieve the goal you state. Being rough with one newbie will not, in my experience, particularly prevent the next question asked by the next newbie :) All IMHO, obviously. Jules

| > Very timely! It's sad that haskell-cafe has so much noise now. | | I disagree with that characterisation. I don't mean to be pedantic, but | I don't think haskell-cafe has lots of noise. I think it has lots of | signal! Quite different. | | We don't have a problem (in my perception, at least) with the quality of | either posts of responses. Not a general one, anyway. The only problem | is that the volume is increasing; which is a problem if that makes it | hard for valuable contributors to keep contributing FWIW I agree with Jules's response here. Yes, the traffic is increasing, but much of it is in long threads with multiple response, and email readers are good at letting you skip that if you don't want to read it. Furthermore, we have the much-lower-bandwidth 'haskell@haskell.org' list, for those who want to see announcements and discussion starters, but don't want the full haskell-cafe experience. We need at least one forum in which it's acceptable to ask anything, no matter how naive, and get polite replies. (RTFM isn't polite; but "The answer is supposed to be documented here (\url); let us know if that doesn't answer your qn" is fine.) I'd be sorry to lose that. Simon

On Fri, 13 Jul 2007, Simon Peyton-Jones wrote:

We need at least one forum in which it's acceptable to ask anything, no matter how naive, and get polite replies. (RTFM isn't polite; but "The answer is supposed to be documented here (\url); let us know if that doesn't answer your qn" is fine.) I'd be sorry to lose that.

Agreed. That said, it might be worth developing a "scary maths FAQ" that's an explanation largely of why all the maths and why it may irritate a lot of people to get in a flap about it all. It seems to come up increasingly often, and... well, it's not just the community being elitist or shutting people out! -- flippa@flippac.org The task of the academic is not to scale great intellectual mountains, but to flatten them.

On Fri, 13 Jul 2007, brad clawsie wrote:

to improve the list, might i suggest

- push chatter to IRC

This is problematic for some kinds of techie chatter, where email makes it easier to get all the maths down.

- take this service off of email entirely. try a web forum system (you may have to slum it and use php). i don't recommend nntp, that just forces us to use gmane since very few isps provide nntp now. a web forum would allow you to segment interest sections while retaining a global search etc. if you use code like slash, you can just moderate noise makers off the page. you can set up a yahoo group in ten minutes.

A global search might be an idea to add to our existing archives, otherwise I'm still not convinced. -- flippa@flippac.org Society does not owe people jobs. Society owes it to itself to find people jobs.

FYI, Gmail *can* kill threads, the Geniuses just deemed it unworthy of a UI presence. This is news to me and related to earlier comments in this thread. HTH http://mail.google.com/support/bin/answer.py?hl=en&answer=47787 On 7/13/07, Nicolas Frisby wrote:

Perhaps an information retrieval pipedream, but what if we attempted an automated FAQ answerer? I'm sure some keywords pop-up often enough in certain chunks of first posts (heterogenous lists, existential error messages, SOE and graphics, category functor monad, etc). It could respond with the standard links to the Wiki pages and research papers.

Of course we would want to keep it on a leash for a while during training and taming.

If we integrate it with the list, it might even belay the list message, giving the user a chance to read its response before selecting: * not quite helpful, send to list anyway * never send me this rubbish again Or we could avoid such invasive methods all together and just auto-post a rather comprehensive answer.

Just a neat thought -- and another potentially nifty and well-sought tool written in Haskell.

On 7/13/07, Philippa Cowderoy wrote:

...

On Fri, 13 Jul 2007, brad clawsie wrote:

...

to improve the list, might i suggest

- push chatter to IRC

This is problematic for some kinds of techie chatter, where email makes it easier to get all the maths down.

...

- take this service off of email entirely. try a web forum system (you may have to slum it and use php). i don't recommend nntp, that just forces us to use gmane since very few isps provide nntp now. a web forum would allow you to segment interest sections while retaining a global search etc. if you use code like slash, you can just moderate noise makers off the page. you can set up a yahoo group in ten minutes.

A global search might be an idea to add to our existing archives, otherwise I'm still not convinced.

-- flippa@flippac.org

Society does not owe people jobs. Society owes it to itself to find people jobs. _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

brad clawsie wrote:

to improve the list, might i suggest

- push chatter to IRC

- take this service off of email entirely. try a web forum system (you may have to slum it and use php). i don't recommend nntp, that just forces us to use gmane since very few isps provide nntp now.

Just curiose, but... what does NNTP have to do with your ISP?

Mark T.B. Carroll wrote:

Andrew Coppin writes:

...

Just curiose, but... what does NNTP have to do with your ISP?

Someone has to provide an NNTP server for your client to connect to. Many ISPs don't. Things like gmane can be used, of course. Some of the free ones can be okay. I like http://news.individual.net/ who, while not free, are pretty cheap, and they carry comp.lang.haskell.

...and when you view a web page, your web browser has to connect to a web server somewhere. I don't see your point...

Mark T.B. Carroll wrote:

Andrew Coppin writes:

...

...and when you view a web page, your web browser has to connect to a web server somewhere.

I don't see your point...

Very many news servers will only serve news to people on the network of whoever's running the server: i.e. a rather restricted 'customer' base.

Really? Most web servers will accept a connection from anybody. (Unless it's *intended* to be an Intranet.) I'm not quite sure why somebody would configure their NNTP server differently...

From: Andrew Coppin

Really? Most web servers will accept a connection from anybody. (Unless it's *intended* to be an Intranet.) I'm not quite sure why somebody would configure their NNTP server differently...

The scale of an NNTP server is simply a *lot* bigger than most web servers, where you only need as much storage capacity and bandwidth as you have content to offer. Daily traffic on the whole of the Usenet takes up a few terabytes, and you presumably want to store more than one day worth of Usenet traffic. You also need to keep in sync with the other Usenet servers, as you are not the sole provider of content. So, NNTP servers need to be powerhouses if they hope to serve even part of the Usenet, and they're closer to search engines than to HTTP servers in terms of the resources required. (For what it's worth, my ISP uses a high-end, dedicated NNTP provider for their Usenet service. For a regular user, it actually costs more to subscribe to that NTTP server alone than to my ISP. Unsurprisingly, my ISP only authorizes its own users to access that particular service.) Anthony Chaumas-Pellet

Anthony Chaumas-Pellet wrote:

From: Andrew Coppin

...

Really? Most web servers will accept a connection from anybody. (Unless it's *intended* to be an Intranet.) I'm not quite sure why somebody would configure their NNTP server differently...

The scale of an NNTP server is simply a *lot* bigger than most web servers, where you only need as much storage capacity and bandwidth as you have content to offer.

Daily traffic on the whole of the Usenet takes up a few terabytes, and you presumably want to store more than one day worth of Usenet traffic. You also need to keep in sync with the other Usenet servers, as you are not the sole provider of content.

So, NNTP servers need to be powerhouses if they hope to serve even part of the Usenet, and they're closer to search engines than to HTTP servers in terms of the resources required.

(For what it's worth, my ISP uses a high-end, dedicated NNTP provider for their Usenet service. For a regular user, it actually costs more to subscribe to that NTTP server alone than to my ISP. Unsurprisingly, my ISP only authorizes its own users to access that particular service.)

What is this "Usenet" thing people keen mentioning? When I want to visit the POV-Ray website, I point my HTTP client at www.povray.org and it displays the page. When I want to use the POV-Ray NNTP forums, I point my NNTP client at news.povray.org and it displays the forums. It works the same way. I don't see how this involves "a few terabytes"...

On Jul 15, 2007, at 0:47 , Jonathan Cast wrote:

Usenet is a giant network of NNTP servers (and UUCP servers before that...) that ISPs (and various Unix sites before that) maintained at one time (most seem to have given up on it now), with thousands of general-purpose newsgroups that at one time were the premier online community.

I'll just point out that NNTP is a *terrible* protocol for news *readers*; it was originally designed to facilitate linking individual news servers. (The various modern NNTP implementations tend to detect readers and switch to a lower overhead NNTP implementation which also restricts some things to keep readers from taking up too much overhead.) Somewhat ironically given the origin of this thread, CMU's solution was to gateway NNTP to IMAP and require readers to use IMAP clients to read the resulting "bboards". -- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH

Perhaps I haven't found the amazing treasure trove of open NNTP servers you appear to have, but in my experience I've yet to find a single good (read: access to most groups and quick about it) and free NNTP server (read: not from my ISP, employer, or university - which, if provided at all, have always been quite limited in which groups they serve), so I can completely understand where the others are coming from in this regard. Perhaps those of you who have found good, free NNTP servers would care to share these well kept secrets? -----Original Message----- From: haskell-cafe-bounces@haskell.org [mailto:haskell-cafe-bounces@haskell.org] On Behalf Of Andrew Coppin Sent: Friday, July 13, 2007 5:04 PM To: haskell-cafe@haskell.org Subject: Re: [Haskell-cafe] Maintaining the community Mark T.B. Carroll wrote:

Andrew Coppin writes:

...

...and when you view a web page, your web browser has to connect to a

...

web server somewhere.

I don't see your point...

Very many news servers will only serve news to people on the network of whoever's running the server: i.e. a rather restricted 'customer' base.

Really? Most web servers will accept a connection from anybody. (Unless it's *intended* to be an Intranet.) I'm not quite sure why somebody would configure their NNTP server differently... _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe -------------------------------------------------------- NOTICE: If received in error, please destroy and notify sender. Sender does not intend to waive confidentiality or privilege. Use of this email is prohibited when received in error.

Hi

I tried an experiment this week of just taking someone's post (Conor's idiom brackets), and putting directly on the wiki first, then letting the author know that's happened.

How do people feel about allowing posts in -cafe to be placed on the wiki, without extensive prior negotiation? What copyright do -cafe@ posts have?

If you email to a public mailing list, you clearly don't have a problem with verbatim copying, since you already know its going to appear in loads of different archives. You also don't have a problem with selected copying, because quoting is how mailing lists work. I'd just assume that all posts to haskell-cafe are "for the good of mankind", and then let them be reused at will.

Should we feel free to put mailing list material onto the wiki?

Yes! Some people do email this list with disclaimers such as "if this email was not specifically addressed to you, and you accidentally remember something from it, we reserve the right to lobotomise you at a future date" - but I think usually those are company policy rather than personal choice. Thanks Neil

On Friday 13 July 2007, Donald Bruce Stewart wrote:

claus.reinke:

...

personally, i tend to be more willing to answer questions on the list than to fiddle with wiki markup and conventions, but there is no reason why people who are happier with wiki editing cannot extract content from list answers to the wiki, especially if its a faq answer rather than a research result.

I've got a few tools that make wiki editing easier (shortcuts to open up a new wiki page for editing in vim, syntax highlighting, console access). These make wiki editing roughly as cheap as composing an email.

I tried an experiment this week of just taking someone's post (Conor's idiom brackets), and putting directly on the wiki first, then letting the author know that's happened.

How do people feel about allowing posts in -cafe to be placed on the wiki, without extensive prior negotiation?

Well, anything /I/ write is OK. . .

What copyright do -cafe@ posts have?

Legally? I'd imagine it's pretty restrictive; morally, I think we should encourage people to wave whatever rights they have (like with the wiki).

If there was a rough consensus that this is ok, we could probably get a lot more material directly on the wiki, since I for one would act first, putting some interesting Clause Reinke posts there semi-verbatim, rather than pondering whether to write an email to the author to seek permission, or cojole them into doing it.

Should we feel free to put mailing list material onto the wiki?

We should. We may not, yet, but that should change. Jonathan Cast http://sourceforge.net/projects/fid-core http://sourceforge.net/projects/fid-emacs

ctm:

On 13 Jul 2007, at 14:47, Donald Bruce Stewart wrote:

...

I tried an experiment this week of just taking someone's post (Conor's idiom brackets), and putting directly on the wiki first, then letting the author know that's happened.

Seemed entirely reasonable to me. If I have a spare moment, I might even finally get around to registering so I can add the other thing I usually bundle (for inserting effectful computations which don't contribute an argument to the main function).

If I'm writing for a mailing list, I'm certainly willing (if surprised) to be exploited in this way.

If there is any legalistic need for explicit permission, then we should have a permanent permission system on an opt-in basis, perhaps recorded in some suitable central and accessible place (like, erm, ...). I'm in.

I've created a page to track contributors who are happy to have their work moved to the Haskell wiki, and thus explicitly licensed under the `simple permissive license'. http://haskell.org/haskellwiki/Haskell_Cafe_migration Just add your names, so your code and text can escape the haskell-cafe copyright monad :) -- Don

...

I've created a page to track contributors who are happy to have their work moved to the Haskell wiki, and thus explicitly licensed under the `simple permissive license'.

http://haskell.org/haskellwiki/Haskell_Cafe_migration

Hi Don, i'm all for using mailinglist postings to improve the wiki, using the wiki as a cache to memoize answers to frequently asked questions, but i have to admit to feeling rather uncomfortable with a general copyright waiver like that. would it mean i'd have to be more careful about what i write on the list (if Oleg wanted to add his name, would he still post those serialized essays)? or would i have to think for each mail whether or not to add a more specific license to disable the default waiver (if you post material you may later want to use in your book, or interim results from your research projects; remember, anything on the wiki is free for all, so anyone could "republish" it if it ends up there..)? if you hadn't told Conor about wikifying his idiom bracket examples, he would not have thought about adding more stuff. nor would he have known that you liked them or that there's now another place where people can find out about, discuss and edit that idea. will everyone start to use those standard disclaimers as signatures by default, in case some code/idea they post might otherwise fall under their copyright waiver (which they signed on to because they thought it was generally a good idea)? why should we have to think about licensing at all? it is nice to see so many of you willing to go that far, but i have been burned before, not on a haskell list, and such a waiver would have left me completely out of the loop. none of the scenarios i can think of would even need such a copyright waiver: 1) someone notices a faq, and decides to add an entry to the faq category on the wiki; they list the most common forms of the question, look through the mailing list archive for some typical good answers, write a little intro and add links to the archived answers. no problem. 2) similar to (1), but none of the existing emails is directly useful without a lot of context from their threads; so, instead of just linking to the emails, someone writes a little text of their own, inserting quotes from some of the archived emails where useful, and adding links to the original emails at the end. or someone takes ideas and quotes from several emails to form a coherent summary of a thread. or someone takes ideas from their specific threads to present a generalized answer, again providing links to the original emails. the only problem i can see here would be misrepresentation by quoting out of context, or editing quotes, but those are standard wiki issues. no new problem. 3) someone writes an entirely new text to answer a faq more directly. no need to reference other texts, unless they were used as the sources of non-standard ideas. no problem. 4) someone writes a text that mixes original contributions with extensive excerpts from other texts, possibly modified; either the plan or the result should be discussed with the original authors, so that they can have a say in how their work is used. in as much as someone else's work is used or modified for the final result, it should be acknowledged, with modifications and permissions made clear. no problem. i am happy for my contributions to this list to be referenced, or quoted with reference, and in most other cases, i'd expect simply to say "go ahead, thanks for doing this" if someone wanted to put in the work to improve the wiki partly based on something i wrote; but unless the parts taken from my emails are trivial, or folklore, or explicitly referenced as separate from the new text, i would expect to be asked, so that i have the chance to say "no" or "not in this way". once i've said "yes", the general wiki license applies, of course. does that sound unreasonable? claus

(sorry if you already know this, just want to clarify. All AIUI, IANAL, etc)

neither am i!-)

If you publish something under licence A, you still remain the copyright holder, and can later also publish it under licence B. You can also publish it combined with other material under licence B.

yes, and nobody is forced to sign on to that waiver list, either. but some of those who do might find it hampering rather than encouraging their own contributions. knowing as they do that someone else might publish their results for and before them, and that they have given full permission for that to happen. i would not like to see list contributions from those active community members falter because of such possible side-effects. most uses do not even seem to require that waiver. and those who do not sign on might still be perfectly happy to respond to most requests with a simple "ok, go ahead".

...

why should we have to think about licensing at all?

If you want code you write to be distributed by Debian, for example, then you need to license it appropriately.

yes, and for the wiki it makes sense, and for open-source projects it makes sense, and i prefer the least limiting licenses whenever possible. i was just pointing out that the default copyright might make more sense for the mailing list, imho. a simple email informing the original authors shouldn't be that hard, should it? jao (just another opinion;-) claus

Ian Lynagh wrote:

If you publish something under licence A, you still remain the copyright holder, and can later also publish it under licence B. You can also publish it combined with other material under licence B.

For example, if you were to write a couple of pages on type systems to the haskell-cafe list, publish it under the "Simple Permissive License", you could still publish a book on Haskell, including your text on type systems, under a more traditional proprietary licence. People would still be able to copy (etc) your type systems text, but would not be able to copy (etc) the other material in your book.

It appears that most publishers want an exclusive license to the published work, i.e. you have to transfer them your copyright. In other words, the publishers won't accept your work if you already published it somewhere else. A Simple Permissive License will most likely disable possible negotiations with the publisher to make them publish anyway in that case. Eric Weisstein's Mathworld has been bitten quite painfully by such copyright transfer, although in the other direction http://mathworld.wolfram.com/about/erics_commentary.html In the end, copyright and laws in general aren't for the case when everybody behaves nicely, but for the case when things go awfully wrong. Regards, apfelmus

Claus Reinke writes:

will ultimately make its contents easier to find. but if you want to avoid answering questions again and again on the list, you need to improve the cache of answers.

Bingo. On less technical forums, e.g. FlyerTalk, the "do a search" equivalent to "RTFM" is rampant and people goal-tend without actually doing a test search to find out if that search might not be coming up for some reason. Others say "do a search" but demonstrate an effective keyword combination and the resulting useful resource. Here, the Wiki is fantastic but extraordinarily spotty (any healthy wiki will always have much new growth, but the current gaps are surprising), and newcomers like myself can and have been contributing to it. If somehow the surge in Haskell Cafe interest could be harnessed to feed Wiki content, that would be great. Something like "I noticed the Wiki entry for list comprehensions is just a stub, so I edited it to answer your question. Is the answer clear? Your beginner perspective is invaluable; how can it be improved?" The only way to establish such a tradition is by example. I'll start giving it a try.

On Fri, Jul 13, 2007 at 08:23:41PM +0100, Andrew Coppin wrote:

Dave Bayer wrote:

...

Here, the Wiki is fantastic but extraordinarily spotty (any healthy wiki will always have much new growth, but the current gaps are surprising), and newcomers like myself can and have been contributing to it.

Whenever I find that something isn't on the Wiki, I try to add it. (E.g., the articles on alpha/beta/eta reduction.)

On the other hand, when I find something isn't there, it's usually because I'm trying to look it up because I don't understand it. ;-) To some extend, incorrect information can be worse than no information... :-(

Conversely, it's generally easier for us to fix an incorrect page than to write a new one from scratch. Stefan

On Fri, 2007-07-13 at 14:33 +0100, Claus Reinke wrote:

...

As we sit here riding the Haskell wave:

http://www.cse.unsw.edu.au/~dons/tmp/cafe.png

with nearly 2000 (!) people reading haskell-cafe@, perhaps its time to think some more about how to build and maintain this lovely Haskell community we have.

- haskell-cafe is meant to be the general forum, that shouldn't change. but i think there is potential to spin off one or two more specialist lists (not too many, or they'll dry out, and not too specific, or they won't attract the haskell-cafe style of membership and content; we also do not want to start cross-postings to keep the synergies of a multitopic forum).

the most obvious one being 'haskell-performance' for shootout entries, 'how do i improve this?', 'what is wrong here?', and 'why isn't haskell slow?' style of questions, profiling, space&time leaks, compiler benchmarks, optimizations, transformations, representations, libs, tools, papers, etc.

I really like this. ^

another possible candidate, judging from mails and blog postings, might be 'haskell-math', for numeric and algebra libs, apps, tools, classes, theory, and math-related algorithms and data structures, and general discussions.

This I'm much much less certain or keen about. Most such questions start as legitimate Haskell questions. Furthermore, I think the replies are often helpful to people who probably wouldn't subscribe to a 'haskell-math' list. (Still it would be nice to have such a venue to just talk about the relation between Haskell and math.) I don't think almost anyone has a problem with such discussions and it seems that many "non-theoretical" readers enjoy them.

On Fri, 2007-13-07 at 20:39 +0200, Andrea Rossato wrote:

...

http://headrush.typepad.com/photos/uncategorized/buildingausercommunity.jpg

Now, I'm telling this because I believe that the expert ones are in part responsible for the gap the picture shows.

In many ways the experts in any such community (not just the Haskell one!) are, in fact, to blame for that gap, yes. The problem is that they, being the experts, can most quickly and effectively answer the questions. And for a while they do (at least subconsciously, but sometimes overtly) bask in the attention they get as the "go-to guy". The problem with this is twofold: 1. As you so ably pointed out, they do a disservice to the intermediates who are sharpening their skills both technical and pedagogical. The intermediates shut up and wait for the experts to answer thus leaving observers of the community thinking there's really only two or three who "really understand" the topic in question. 2. The experts wear out. Eventually the attention isn't something to bask in. It's to be avoided. The result is the experts avoiding places where they're likely to be prevailed upon to answer questions and, because of the training from #1, nobody is left behind to fill the resulting gap. I've seen this pattern so often in communities. I've also seen it in management (the supervisor/manager who can do the job better than his underlings -- so he does) or in teaching (the popular teacher gets a heavier courseload, in effect being punished for being good) or in a myriad of other social enterprises. -- Michael T. Richter (GoogleTalk: ttmrichter@gmail.com) A well-designed and humane interface does not need to be split into beginner and expert subsystems. (Jef Raskin)

On Sun, 2007-15-07 at 10:56 +0200, Andrea Rossato wrote:

...

I've seen this pattern so often in communities.

I may be wrong, but I think you do not get the specificity of the Haskell community, that is quite peculiar, I'd say.

I think you're wrong. ;) The specifics of motivation and style will vary from community to community, but the net result is alarmingly (and sometimes depressingly) the same: newcomers are driven away just at the point where they're transitioning from beginners who take from the community to intermediates who could (in theory) contribute. The specifics in the Haskell community are more academic in their nature because, as you pointed out, most of the grognards are people who live and breathe academia. The Python community has a different engine for the dynamic but very similar results (right down to the mockery when people think they're not being listened in on). The Linux kernel team is infamous for its hostility to outsider input and has driven off a lot of talent that could have contributed greatly to fixing up its internal nightmares. The FSF is a personality cult with all the negatives that entails, again with people leaving just at the point where they could become useful contributors. The list goes on and on. One of the frustrating things about seeing this happen over and over again is the insistence of people that "this community is somehow different". It isn't. People are people and politics is politics. It's all the same manure on a different pile.

The only reaction I received was a couple of Haskell Top Gurus making fun of me in the #haskell IRC channel (they did not know I was reading, probably). As a reaction I just wanted to erase my wiki pages and quit the community. I did only the later, but the tutorial remains unfinished (I was thinking to finish it, but read on).

I was always wondering what happened to your tutorial. Now I know. The mockery on a public forum? Was not cool. Hell, mockery with other tops over private channels would be a childish and unhelpful reaction, but making it open like that is outright destructive. I'm quite frankly disappointed to hear that it happened.

And even I'm doing my academic career in Italy, known to have a very corrupted academic system [...]

I'll pit China against Italy for corruption any day. ;)

I don't like speaking bad about a community I'm not part of.

On the one hand I want to say "please speak as badly as you can as often as you can" because the only way that problems that are driving off potentially productive members of the community can be fixed is if they're stared at in the face and hit with the rolled-up newspaper of decency. "Bad dog!" On the other hand, there's this: http://www.shirky.com/writings/group_enemy.html It makes me wonder if there's any hope at all for online communities. -- Michael T. Richter (GoogleTalk: ttmrichter@gmail.com) The most exciting phrase to hear in science - the one that heralds new discoveries - is not "Eureka!" but "That's funny..." (Isaac Asimov)

Andrea Rossato wrote:

I may be wrong, but I think you do not get the specificity of the Haskell community, that is quite peculiar, I'd say.

Interesting. I've just read your message three times, and you argue eloquently, and coherently. But I feel a bit of a disconnection. I don't recognise the community you're describing! I've found the haskell community to be one of the very friendliest I've ever come across on the internet. Not just friendly to me, but friendly to everyone who comes along. Every time I go on IRC, I see people (complete beginners, or less so) receiving a standard of free help and tuition which would put many of the world's leading computer training courses and degree courses to shame. (I also don't recognise your description of the academic community; anecdote vs anecdote but your experience was clearly quite different from mine) Jules

This is a bit tangental, but... One problem I sometimes have is not knowing the status of things. E.g., you read about Associated Types, and then you go "hey, is this implemented now? is it being implemented soon? etc."

(Don't all rush in and tell me about ATs - I'm only picking it as an example. Others include, say, the GHC debugger, stream fusion, parallel arrays, etc etc etc.)

Maybe I'm looking wrong, but it often isn't obvious to me how to figure out the answer to the question "does all this cool stuff work yet?" I'm not really sure what we could change to fix that though...

the Haskell Community & Activity Reports were created in part to answer that kind of question. within the limitations of 6-monthly editions (too frequent for some projects, too infrequent for others), they have tried to do ever since November 2001, the latest dating from May 2007: http://www.haskell.org/communities/ since contributors are encouraged to look ahead in plans, as well as look back on achievements, the interval of not knowing where things stand is actually shorter than 6 months. provided that people remember to write and submit entries about their favourite haskell projects when the editor comes calling before November/May!-) claus

On Sun, 2007-07-15 at 17:49 +0100, Andrew Coppin wrote:

This is a bit tangental, but... One problem I sometimes have is not knowing the status of things. E.g., you read about Associated Types, and then you go "hey, is this implemented now? is it being implemented soon? etc."

(Don't all rush in and tell me about ATs - I'm only picking it as an example. Others include, say, the GHC debugger, stream fusion, parallel arrays, etc etc etc.)

Maybe I'm looking wrong, but it often isn't obvious to me how to figure out the answer to the question "does all this cool stuff work yet?" I'm not really sure what we could change to fix that though...

Ah, a question that can be answered with RTFM (http://haskell.org/haskellwiki/GHC)

On Sun, 2007-07-15 at 18:47 +0100, Andrew Coppin wrote:

Derek Elkins wrote:

...

On Sun, 2007-07-15 at 17:49 +0100, Andrew Coppin wrote:

...

This is a bit tangental, but... One problem I sometimes have is not knowing the status of things. E.g., you read about Associated Types, and then you go "hey, is this implemented now? is it being implemented soon? etc."

(Don't all rush in and tell me about ATs - I'm only picking it as an example. Others include, say, the GHC debugger, stream fusion, parallel arrays, etc etc etc.)

Maybe I'm looking wrong, but it often isn't obvious to me how to figure out the answer to the question "does all this cool stuff work yet?" I'm not really sure what we could change to fix that though...

Ah, a question that can be answered with RTFM (http://haskell.org/haskellwiki/GHC)

I could have sworn I heard somewhere that the debugger would be in GHC 6.6.1... but, apparently, I am mistaken. I have no idea which version it *is* going to be in... (Presumably the next one.)

On the very page I linked is a link labelled "The GHCi Debugger". Clicking on it you get "The debugger is currently available in GHC 6.7" Similarly for parallel arrays There is no version of bytestrings without stream fusion and there never was. Bytestrings have no compiler support, it is just a library.

On 7/15/07, Derek Elkins wrote:

There is no version of bytestrings without stream fusion and there never was. Bytestrings have no compiler support, it is just a library.

I'm not sure that's correct. Stream fusion is a particular fusion technique that wasn't introduced until fairly recently.

From what I can tell, none of the versions available from http://www.cse.unsw.edu.au/~dons/fps.html include it. You have to go to http://www.cse.unsw.edu.au/~dons/papers/CSL06.html and get the code from the fps-unstable branch.

Donald Bruce Stewart wrote:

zednenem:

...

On 7/15/07, Derek Elkins wrote:

...

There is no version of bytestrings without stream fusion and there never was. Bytestrings have no compiler support, it is just a library.

I'm not sure that's correct. Stream fusion is a particular fusion technique that wasn't introduced until fairly recently.

...

From what I can tell, none of the versions available from http://www.cse.unsw.edu.au/~dons/fps.html include it. You have to go to http://www.cse.unsw.edu.au/~dons/papers/CSL06.html and get the code from the fps-unstable branch.

That's right. Both stream fusion for lists and bytestrings are currently only in darcs,

http://www.cse.unsw.edu.au/~dons/code/fps-unstable/ http://www.cse.unsw.edu.au/~dons/code/streams/list/

Bytestrings will be streamed by the next release.

...which brings me back to my original "how do I know if it's there?" question. ;-) My copy of GHC sitting here certainly *has* support for lists and byte strings in it - but I have no clue what version...

"Alex Queiroz" wrote:

This is so much true. It has the effect of disguising Haskell as a PhD-only language.

And what would be wrong with Haskell being a PhD-only language, if it were true? OK, so I'm not genuinely suggesting that you must possess or be studying for a PhD, to grok Haskell. But I find nothing alarming about the suggestion that one needs a fairly high level of intelligence, and some training, in order to be able to use Haskell effectively. After all, we would expect the same attributes (intelligence and training) from a neurosurgeon, a nuclear scientist, or someone who calculates how to land a person on the moon. Programming computers may not seem very skilled to most people, but maybe that is simply because we are so familiar with it being done so badly. I'm all for improving the quality of software, and the corollary is that that means improving the quality of programmers (by stretching our brains!). Regards, Malcolm

Alex Queiroz wrote:

On 7/16/07, Malcolm Wallace wrote:

...

OK, so I'm not genuinely suggesting that you must possess or be studying for a PhD, to grok Haskell. But I find nothing alarming about the suggestion that one needs a fairly high level of intelligence, and some training, in order to be able to use Haskell effectively.

What I'm saying is that almost every topic in Haskell Café evolves into a very high level discussion that may frighten some beginners, as it seems that without a PhD in programming languages and category theory, the language is not for you.

read . takeWhile (not . frightening) ;) Personally, I perceive Haskell as being easier than every other programming language. In other words, if Haskell requires a PhD, Visual Basic requires a Nobel Prize. How the heck do imperative programmers produce working code and how are they able to read the resulting mess afterwards? I just don't get it :) To be serious, those frightening things are often very simple concepts but will remain frightening if not explained well. My experience is that wikis, blog posts and online tutorials can't replace a textbook-quality, well, textbook. Unless the online materials are textbook-quality as well, of course. Really, the best way to learn Haskell (and most other things) is to read/buy/borrow a textbook. This also applies to the mailing list and the "cache of answers" for optimization volume. One example is the "hGetContents - hClose" question. I think that most people encountering this problem won't realize on the first try that hGetContents is the culprit. But how to formulate a good search query then? In the end, I think that the best way to avoid trouble with hGetContents is to be introduced to it in a textbook chapter "IO and Files". Regards, apfelmus PS: hGetContents-hClose is particularly strange since you need operational semantics of lazy evaluation to understand it.

On 7/16/07, Malcolm Wallace wrote:

After all, we would expect the same attributes (intelligence and training) from a neurosurgeon, a nuclear scientist, or someone who calculates how to land a person on the moon. Programming computers may not seem very skilled to most people, but maybe that is simply because we are so familiar with it being done so badly. I'm all for improving the quality of software, and the corollary is that that means improving the quality of programmers (by stretching our brains!).

You want people doing difficult expensive high-risk tasks to be intelligent and well trained, but you want their task to be as easy as possible. Would you rather a nuclear reactor needs to be controlled by feeding in punch cards, or by having a big round dial labelled "power", that you can move from 0 to 200 MegaWatts? Of course, you'd like the guy moving that dial to be well trained and intelligent. Welcome to why flying airlines is well-paid and boring.

On 7/16/07, Malcolm Wallace wrote:

OK, so I'm not genuinely suggesting that you must possess or be studying for a PhD, to grok Haskell. But I find nothing alarming about the suggestion that one needs a fairly high level of intelligence, and some training, in order to be able to use Haskell effectively.

When I was a teenager I thought people with PhDs were minor deities. Having done one, and knowing lots of people with them, I can tell you, the vast majority of people with a PhD (including me) have merely above average intelligence. A PhD is not a mark of intelligence. It's a mark of persistence. (Shall we say obsession?) I think Malcolm's analogy to other professions is quite apt. If we expect to be taken seriously as professionals, it would be unsurprising to find that we need to engage in some strenuous [mental] effort to acquire the skills. And this is where I think Haskell has it all over C++, Java, and the rest. Haskell is easy to learn at a simple level, and hard to learn at the expert level, but once learned is very powerful and has excellent payoffs in terms of productivity. With C++ or Java, the expertise is somewhat easier to acquire, but you never get the payoff. And before you all flame, yes, I do know C++ at an expert level, and that is exactly why, after 7 years of writing server software in C++, I now want to do it in Haskell. cheers, T -- Dr Thomas Conway drtomc@gmail.com Silence is the perfectest herald of joy: I were but little happy, if I could say how much.

On 7/17/07, Martin Coxall wrote:

Me too, which is why I find your statement that expertise in C++ is easy to acquire. Seeing some of my colleagues' code is enough to tell me that this is most definitely not the case.

You're quite right. That was careless on my part. Though the way C++ is taught at the undergraduate level, and the way it is perceived by the inexperienced is that it isn't so hard. But then again, I've taught Java at the ugrad level, and what do I know about Java, other that it'd be quite a nice place to have a holiday some time.... :) T. -- Dr Thomas Conway drtomc@gmail.com Silence is the perfectest herald of joy: I were but little happy, if I could say how much.

On 7/17/07, Magnus Therning wrote:

On Tue, Jul 17, 2007 at 19:43:51 +1000, Thomas Conway wrote:

...

On 7/17/07, Martin Coxall wrote:

...

Me too, which is why I find your statement that expertise in C++ is easy to acquire. Seeing some of my colleagues' code is enough to tell me that this is most definitely not the case.

You're quite right. That was careless on my part. Though the way C++ is taught at the undergraduate level, and the way it is perceived by the inexperienced is that it isn't so hard.

But then again, I've taught Java at the ugrad level, and what do I know about Java, other that it'd be quite a nice place to have a holiday some time.... :)

My wife is from Java so if you need some advice on good vacation spots I'd be happy to ask her advice ;-)

import java.wife.advice.CheapHotels;

On 7/18/07, Hugh Perkins wrote:

Am I the only person who finds it interesting/worrying that there are few to no people in the group who are ex-C# programmers. I mean, you could argue that C# programmers are simply too stupid to do Haskell, but ... you know, there is another explanation ;-)

I wouldn't say too stupid, but it may be a cultural thing. People working in C++ are more likely to be doing what I would call "technical" programming, and correspondingly more likely to be interested in Haskell, and to appreciate what it has to offer from painful personal experience. From what I know of the marketplace, people working in C# are more likely to be doing client/integration work where technical finesse is less important, and are therefore less likely to see the point. cheers, T. -- Dr Thomas Conway drtomc@gmail.com Silence is the perfectest herald of joy: I were but little happy, if I could say how much.

On Tuesday 17 July 2007 23:26:08 Hugh Perkins wrote:

Am I the only person who finds it interesting/worrying that there are few to no people in the group who are ex-C# programmers. I mean, you could argue that C# programmers are simply too stupid to do Haskell, but ... you know, there is another explanation ;-)

To understand this, I think you must look at the number of technical users for each language. There are a huge number of technical C++ and Java programmers but a tiny number of technical C# programmers in comparison. The few technical C# programmers are migrating to F# because it is next door and F# programmers are better looking. You can find evidence of this simply by searching for mundane numerical libraries like Fast Fourier Transform (FFT) implementations. If you do this for C++ or Java you find hundreds of implementations, some of which will work. For OCaml, you find a handful of libraries all bundled with automated proofs of their correctness. When I did this for all .NET languages a couple of months ago, there was nothing worth having (even among expensive commercial solutions). So I wrote my own and productized it. Our FFT library in C# is an order of magnitude less popular than our technical libraries for OCaml but this is offset by the fact that people using C# have an order of magnitude more money. So I would say that the Haskell community can expect immigrants who are technical (or they wouldn't consider a fringe language) and that means they will be migrating primarily from C++ and Java. If you want to attract the maximum number of users then target your educational material at those people. The C++ programmers will know that coding to the metal is always essential and will demand proof that Haskell is as fast as C. They will also need to know how to unravel low-dimensional vector and matrix routines at compile time. The Java programmers will know that performance (particularly startup time) is completely irrelevant but being cross-platform and having extreme interoperability is pivotal. They will be particularly impressed by Haskell's brevity and the disappearance of some of Java's keywords like public, static, void and Factory. -- Dr Jon D Harrop, Flying Frog Consultancy Ltd. OCaml for Scientists http://www.ffconsultancy.com/products/ocaml_for_scientists/?e

On 7/18/07, Martin Coxall wrote:

On 7/18/07, Jon Harrop wrote:

...

On Tuesday 17 July 2007 23:26:08 Hugh Perkins wrote:

...

Am I the only person who finds it interesting/worrying that there are few to no people in the group who are ex-C# programmers. I mean, you could argue that C# programmers are simply too stupid to do Haskell, but ... you know, there is another explanation ;-)

To understand this, I think you must look at the number of technical users for each language. There are a huge number of technical C++ and Java programmers but a tiny number of technical C# programmers in comparison. The few technical C# programmers are migrating to F# because it is next door and F# programmers are better looking.

Most C# programmers are (a) GUI programmers and (b) former VB programmers. This means they are *guaranteed* to be less attractive that the average C++ developer. I have proof. But it's too big to be contained in this margin.

Martin

I heard that Fermat didn't even actually have a proof. You wouldn't be trying to hoodwink us in the same way, would you? :) I haven't been paying attention to the subject, but I suppose I should pipe in now. I really enjoy Haskell. I'm probably like most people here in that I like learning new languages: I've given Scheme a fair shot; F# captured my interest for a while, and right now I'm toying with Erlang. I've tried Python, used Perl for a job, determined after an hour that PHP wasn't for me, and even looked at Ruby. The list goes on. (Always, of course, I keep GHC on my computer.) But for work, I use C#. And I, for one, am looking forward to C#3.0, because it will be easier to apply my FP experience to problems when FP is the better way to solve a problem. (You've heard the maxim that when all you have is a hammer, everything looks like a nail; the flip side of it is that when you've got a whole tool set including a screwdriver and you see a screw, but your company only lets you use your hammer, it can be frustrating to beat on the screw with the hammer.) And since I'm fresh out of college with no experience, I'm neither in a position to even suggest a language change in my company, nor do I have the experience to move to the occassional Scheme or Erlang job opening I see (I don't know if I've ever seen a Haskell job opening, and I'm guessing if I did it would get snatched up by a more qualified programmer quite quickly). I guess the point being made is that there are a smaller percentage of attractive programmers in C#; but it looked to me that people were implying that there are /no/ knowledgeable programmers in C#; and I'd just like to assert that maybe there are some that don't really have a choice right now. :) Bryan

Just found this in a gmail adtext link, it's quite interesting (and convincing): http://www.janestcapital.com/yaron_minsky-cufp_2006.pdf