September 2015 - Haskell-community

Request for comment: New haskell.org download page
by John Wiegley 12 Jul '16

12 Jul '16

As mentioned earlier, I've been working on a draft version of the new Haskell download page in consultation with Simon PJ, Michael Snoyman, and Gershom Bazerman. The goal has been twofold: a) add stack as an explicit option, and b) add text to each option indicating clearly what it provides and where to get further help, so users can understand the options and make an informed choice. We've sought to keep the text factual, rather than imply that one option is "best" for any particular class of user, since opinions vary so widely on this point. At the following link, you'll find a draft version of the new page for comment: https://gist.github.com/jwiegley/153d968ddfc9046ee4c9 Hopefully it can go live on haskell.org next week, so please contribute your edits here, or by pull request. The goal is to explain each option so that people can make an informed decision. However, the order of presentation does imply that whatever comes first is "preferred" even if that is not the intent. The order currently given is HP, Stack, Minimal. Chris has already made a few points about changing this order, so let's continue that discussion and see where it leads us. Bear in mind that this is (hopefully) only an interim state. The plan is to add Stack to the Platform, and render the Platform minimal, which will consolidate this page down to a single, recommended download path. At the bottom of the gist are incomplete sections on third party libraries and alternate installation approaches. These have yet to be written. The hope is to resolve the top content first and sort the rest out after; however, ideas for that content is most welcome too. Thank you, John Wiegley

21 50

Request for comment: New haskell.org download page
by 韩冬 24 Sep '15

24 Sep '15

As a newcomer(start learning about 3 months ago), I also will stand up and say i use the HP, i did meet some problems at the beginning, but the problems are all solved by reading documents and understanding more about ghc package management. Now i’m playing well with cabal sandbox and never had any problems. The main reason to use HP is that HP make ghci usable(i mean fire ghci anywhere, not cabal repl inside a project), otherwise i have to install lots of global packages by my self. Cabal and cabal-install do need some document care, but that’s all, global package doesn’t harm as long as you use sandbox. Finally, i’d say i really wish the minimal installer comes first with document link, that’s the right way to let newcomer understand how the ghc package management works and how to choose between stack/cabal.

1 0

Request for comment: New haskell.org download page
by MightyByte 24 Sep '15

24 Sep '15

On Thu, Sep 24, 2015 at 8:23 PM, John Wiegley <johnw at newartisans.com> wrote: > > Perhaps we need a poll. So far the presented options are: > > 1. HP Stack Minimal > 2. Stack HP Minimal > 3. Stack Minimal HP > 4. Minimal HP Stack > 5. Minimal Stack HP I vote for #4.

1 0

Haskell Platform as the default recommendation considered harmful
by Christopher Allen 21 Sep '15

21 Sep '15

Here are some quotes from end-users regarding Platform. This is just what I found on the web and from skimming IRC logs, it doesn't include many hundreds more of examples from IRC of new people _in particular_ getting stuck because they didn't understand they were conflicting with globally installed packages. New people do not understand: 1. ghc-pkg 2. The global vs. user package databases 3. Cabal 4. What Platform does to their build environment New Haskell users are least equipped to understand errors caused by: 1. Package version conflicts 2. Something innate to their install, rather than something wrong with their package constraints. Part of the problem is that the people making decisions on these matters are not the people on the frontlines helping new people in IRC or otherwise in the wild with sufficient regularity to see how often this catches people up. It's less common than it used to be, but that's due in large part to people following the install instructions on https://github.com/bitemyapp/learnhaskell and using things like MinGHC and Stack. Comments like this from the committee: "Despite the controversies over if platform is "right for beginners" it needs to be provided from the downloads page generally." Are representative of the experience the community has had trying to get the situation rectified and people routed to alternatives so that unnecessary friction isn't encountered. Past comments I've made on this: Haskell libraries typically move pretty quickly...Having cabal sandbox init && cabal install $package just work is important and Platform compromises that as the globally installed packages becoming out of date between releases. Doug Beardsley, one of the core Snap developers said this in reply: "This. This is the biggest reason I don't use the platform and have advised a number of relative newcomers to switch from the platform to straight GHC when they encountered build dependency problems." I used to tell people in my guide to install Haskell Platform, but I stopped because it kept creating problems. Haskell Platform causes trouble for new people when the globally installed packages get old. These very same new people are the ones least equipped to cope with or understand the problem or where it comes from. Platform should be recommended when it is the best option available for the majority of Haskell users. Quotes from end-users: "The HP downloads have been moved "above the fold" with a clear demarcating line separating the "first class citizen" (HP) from the other, minimal installers. Frankly, I was astonished when I went on the other day and saw this change, since originally there was pushback about even having HP on the downloads page at all!" (re: content changes on the new haskell.org website) "I know I'm a beginner, but I don't even understand what problem #Haskell platform is trying to solve." "Yeah, I was bitten by Haskell Platform a few weeks ago. I'll have to check out Stack though." "I haven't used haskell for a few years, installed haskell platform the other day, it works fine but I can't cabal install" (said 2015-09-15 btw) > As per #55 we want less biased presentation of bare install vs. platform, not more. "Ok. (I recommend people to Haskell sometimes, and I see HP hurts them a lot, as they all want to do webdev) Especially now that Stack is out and picking up momentum, I personally consider HP completely superseded." "Windows user here. I dumped Platform the day MinGHC happened, and have never looked back." "On OS X, in my experience it is way easier to set up Haskell without HP than with, fwiw. I don't think it was always this way, but it has been for at least a year." "Actually, I don't understand why we need Haskell Platform at all? Now, when stack exists, Haskell Platform is useless, I think. "Haskell with batteries included"? Yes, but what batteries are? Compiler+Interpreter, Cabal 35 "core" packages. With stack I can setup ghc-infrastructure and install any packages I need." "It seems to me that within the community, we all know stack is the clear winner, and we can recommend it to whomever is showing us an interest in Haskell. Unfortunately, haskell.org appears to be the landing portal for the language to a newcomer with no existing ties to the community, and thus no one to recommend stack over it. Stack is so obviously a better tool for a new user, it's unbelievable. /subjective opinion" "Yes, a regular frustration when helping folks on IRC is that the first three steps have so often been to, 1. Apologize 2. Instruct how to uninstall what they have 3. Start again with GHC and cabal The effort made at packaging GHC on both Windows and OS X are really solid pieces of work. It looks like Stackage LTS will provide a nice basis for a set of libraries moving forward." "as a newborn Haskeller, I agree. Platform (OSX) was a huge pain, uninstalled it, learned cabal sandboxes, sometimes nuke them, everything works." "I find the platform to be a huge burden to beginners on OS X. It just leads to cabal hell. Anything except the minimum to build cabal-install in the global sandbox == super bad." "The platform was very confusing to me as a beginner, and only now that I know what I'm doing do I see what it was trying to accomplish. I'm depending on the awesome PPA, hopefully GHC can be made easier to get running without such a thing on other platforms." "Actually there is a disadvantage with the HP, as I pointed out already in an email-response, as the extra HP packages get registered into the global package db, and therefore pollute the sandbox environment with those extra packages." "The problem is that Cabal prefers the version already installed (i.e. exposed via the global pkg db), when I want the Cabal solver to be as unrestricted as possible in a sandbox in order to get the latest available versions rather than those imposed by the HP." "So is there really a simple way (as you seem to suggest) to hide all the extra HP-provided package versions inside a sandbox?" Recommended solution: > add an entry like this to the constraints field in the cabal.config file in that sandbox: > constraints: attoparsec source "Funny because I have been using the Haskell ecosystem since roughly one or two GHC versions longer than the platform has been around and to me it always felt outdated on average over the lifetime of one HP release. GHC 6.12.1 was released in December 2009, the next platform Beta release was in March 2010 and the next actual release in July 2010. So the platform contained the outdated GHC 6.10 for the better part of a year after the new GHC version was released. GHC 7.0.1 was released in November 2010, again we had to wait until March for a matching platform release. GHC 7.2.1 was release in August 2011 but was marked as a technology preview and completely ignored by the platform. GHC 7.4.1 was released in February 2012, the platform release including it was in June, coinciding almost exactly with the GHC 7.4.2 release which was then ignored by the platform until November, after even GHC 7.6.1 had been released in September 2012 already. To get GHC 7.6 in a platform release we had to wait until May 2013 when the reasonably current (release in April 2013) GHC 7.6.3 was included in the only platform release of that year. In April 2014 we got GHC 7.8.1 and in August 2014 we got a platform release for it which included 7.8.3, the current GHC release at the time and until GHC 7.8.4 was released in December. Not to mention that the HP is not even handled like a regular stable distribution, e.g. it doesn't even update patch levels of packages or compilers." "Haskell does still have major "ease of installation" problems in my experience." "using Haskell Platform at all might cause troubles; it's explicitly recommended against [...] because it uses the global package database." "Haskell Platform feels a bit like Prelude; fine for people who are just using it for small self-teaching projects, but once you're supposed to do something "serious" then people seem to throw up their hands and say "No! never mind that.". I don't know. Maybe I've got the wrong impression. Now that I know enough cabal to get by and not screw myself over (by creating dependency hell), I feel like Haskell Platform might not be relevant to my needs any more." "No, you're perfectly right. And while HP may be up to date right now, it'll be out of date next month and it will keep getting more and more obsolete until the next release in a year or so." "It's still bad and cabal is still bad. I was in #haskell on IRC not too long ago and they couldn't even figure out why a haskell-platform install couldn't upgrade its own cabal with no other packages installed." (Example of Cabal getting blamed for problems with Platform) "If you're on OS X or Linux, you should not use the Haskell Platform. It's a whole lot of pain for literally no gain..." "The Haskell Platform was an attempt to define one but many don't like to depend on it." "I suggest you to use hvr's GHC packages. That way you can simply apt-get install ghc-7.10.1 cabal-install-1.22" "What is the best way of installing GHC on OS X? There's Haskell Platform, Homebrew GHC/Cabal formulas, andHaskell for OS X. I've been told that Homebrew formulas were broken a few months ago, and that Haskell Platform is not good if you plan on doing anything but the most basic things." Example links: https://www.reddit.com/r/haskell/comments/390msw/unable_to_install_ghc_and_… https://www.reddit.com/r/haskell/comments/390msw/unable_to_install_ghc_and_… https://www.reddit.com/r/haskell/comments/390msw/unable_to_install_ghc_and_… https://github.com/haskell/haskell-platform/issues/165 https://www.reddit.com/r/haskell/comments/3b1yuk/haskellinfrastructure_fwd_… https://www.reddit.com/r/haskell/comments/3b1yuk/haskellinfrastructure_fwd_… https://www.reddit.com/r/haskell/comments/3b1yuk/haskellinfrastructure_fwd_… https://www.reddit.com/r/haskell/comments/3b1yuk/haskellinfrastructure_fwd_… https://www.reddit.com/r/haskell/comments/2zts44/wither_the_platform/cpm8okr https://www.reddit.com/r/haskell/comments/2zts44/wither_the_platform/cpmk4vt https://www.reddit.com/r/haskell/comments/2zts44/wither_the_platform/cpmclvo https://www.reddit.com/r/haskell/comments/2zts44/wither_the_platform/cpm8l4a https://www.reddit.com/r/haskell/comments/2zts44/wither_the_platform/cpmbopv https://www.reddit.com/r/haskell/comments/2zts44/wither_the_platform/cpmljeh https://www.reddit.com/r/haskell/comments/2w3php/venerable_haskell_platform… https://www.reddit.com/r/haskell/comments/2w3php/venerable_haskell_platform… https://www.reddit.com/r/haskell/comments/2w3php/venerable_haskell_platform… https://www.reddit.com/r/haskell/comments/2vy1lw/new_haskell_homepage_is_li… https://www.reddit.com/r/haskell/comments/2vy1lw/new_haskell_homepage_is_li… https://www.reddit.com/r/haskell/comments/2w3php/venerable_haskell_platform… https://www.reddit.com/r/haskell/comments/38br54/help_wanted_for_haskell_pl… https://www.reddit.com/r/haskell/comments/38br54/help_wanted_for_haskell_pl… https://news.ycombinator.com/item?id=8083705 https://news.ycombinator.com/item?id=6300091 https://news.ycombinator.com/item?id=9808260 http://www.davesquared.net/2014/05/platformless-haskell.html http://stackoverflow.com/questions/32532120/what-is-the-haskell-standard-li… http://stackoverflow.com/questions/32126861/bytestring-types-fail-when-inst… https://www.reddit.com/r/haskell/comments/3fqiff/dont_give_up_on_haskell/ct… https://github.com/haskell-infra/hl/issues/93#issuecomment-75862519 https://github.com/haskell-infra/hl/issues/55 http://ircbrowse.net/browse/haskell?id=21305217&timestamp=1440372668#t14403… https://twitter.com/mac10688/status/644600976897347584 https://twitter.com/JulianBirch/status/640525865793810432 https://twitter.com/dshevchenko_biz/status/640416836694245376 https://github.com/haskell-infra/hl/pull/130 Seem like a long list? Welcome to the bloodbath. I urge you try to find a similar level of discontent in other programming language communities with their community infrastructure like this. My preference? Stop linking people to Platform until, at minimum, the globally installed packages are removed per: https://github.com/haskell/haskell-platform/issues/206#issuecomment-1383344… -- Chris Allen Currently working on http://haskellbook.com

5 10

Hackage Roots of Trust
by Boespflug, Mathieu 21 Sep '15

21 Sep '15

GershomB wrote: > One related discussion we started to have was if we might want > to do haskell community funding for "phase two" of the update > framework rollout, as discussed in Duncan's talk -- that phase > where we not only implement server trust and signing, but also > author signing. This sounds like a project that potentially the community could consider valuable. However, how much total effort are we talking about here? How much funding would be required? PS: apologies for the top-post. Just signed up to the mailing list and didn't find a way to convince my mail client to set the in-reply-to field appropriately to reference the original message.

1 0

Hackage Roots of Trust
by Gershom B 18 Sep '15

18 Sep '15

At the Haskell Implementor's Workshop at ICFP, Duncan gave a talk on the work on security and package infrastructure that has been going on: https://www.youtube.com/watch?v=D9juHHlnayI One element of that, which was turned over the committee to figure out is who our actual roots of trust would be, in the same sense that there are root certificates for TLS and https authentication, etc. at the Haskell Symposium itself, I gave a quick lightning talk on the work the committee had done in this regard: https://www.youtube.com/watch?v=U8ISiSXV2c0 (If you are interested in verifying your communications with Duncan by the way, and if you trust the video is undoctored, then his GPG key fingerprint appears on it, which may be of some use.) We did in fact get some keysigning done at the conference, and we also secured a fair number of keys from the roots of trust we co-ordinated, though some followup work remains to be done there. We certainly already have enough in hand to bootstrap the process as the hackage security work gets fully rolled out. One related discussion we started to have was if we might want to do haskell community funding for "phase two" of the update framework rollout, as discussed in Duncan's talk -- that phase where we not only implement server trust and signing, but also author signing. Apropos of nothing, but a related thought/question I had was if there would be interest in making cabal files themselves more potentially secure in the manner of the LIO / HLIO work [1]. Having a better chain of trust seems to somewhat obviate the need here, but it does seem like something worth considering. Similar mechanisms might also be worth integrating into template haskell IO for that matter. However, one concern is that the worth of these approaches depends in part on good SafeHaskell takeup, which has a whole bunch of obstacles in itself :-) Cheers, Gershom [1] http://www.cse.chalmers.se/~russo/publications_files/hybrid-icfp2015.pdf and https://hackage.haskell.org/package/lio-0.11.5.0 and http://www.scs.stanford.edu/~deian/pubs/stefan:2014:building-haskell.pdf

3 4

Code signing for Windows exes
by Randy Polen 17 Sep '15

17 Sep '15

It would also be great, while considering the roots of trust, to also consider acquiring and escrowing (i.e., who keeps, who uses) a code signing certificate so that we can sign the handful of Haskell Community executables that we build for Windows. Top of that list would be the Haskell Platform installation executables, but we could also use it for the pre-built cabal, stack, etc.

2 1

Better Stats
by Gershom B 09 Sep '15

09 Sep '15

One of the things I mentioned wanting help with in the haskell infra status update I gave at HIW [1] was better stats. In fact, when I prepared the talk, I forgot we actually had some google analytics in place. In any case, they need work to be extended and made more complete with regards to pages covered, and especially downloads. Jose Calderon volunteered afterwards to put some effort in on this, as time permits. He says he’d like to be able to stamp out “quarterly traffic reports” or the like eventually, which sounds pretty neat to me, and I got him set up (I think) on GA. Again, just recording this here and ideally we can track this as time goes forward. —Gershom [1] https://www.youtube.com/watch?v=cblJf84nkj8

1 0

A few kickoff notes and Outreachy
by Gershom B 09 Sep '15

09 Sep '15

Hi everyone and welcome. I’m surprised we have roughly 50 subscribers so far. Bear in mind I do expect this to be a relatively boring list, but we will of course see how things develop :-). I came back from ICFP with a bunch of discussions about various small things in various stages of work and repair, and I plan to start a number of small threads as time permits over the next few days on each of these in turn, with the idea that people who are interested can follow only relevant discussions, and people can reply over time (be it rapidly or just a month or two down the line) to update on how things are going. Many of these were discussed by the whole committee at one of our first in-person meetings since inception where all members were present (a rare occasion, and not one I think we should imagine is easily replicable). One thing thought I would mention first is that we are planning to participate in Outreachy (https://www.gnome.org/outreachy/) which is the successor to the Gnome Outreach Program for Women. It works sort of like GSoC but with an outreach focus, and with orgs funding their own interns and Outreachy coordinating, as I understand it. We had meant to do so for a few years but the timing hasn’t worked out. The applications for orgs to participate in the new round closes on September 21 — I think we had thought it was too late for us to co-ordinate for this round, but instead we would prefer to do it in May so as to co-ordinate our admin overhead with that of dealing with GSoC. Nonetheless, I believe the plan was to get in touch with Outreachy very soon, so that we could actually have things lined up by the next round, unlike previous years. I think Adam was going to handle this? I think an overall consideration here is that we have now acquired a relative surplus compared to our income, year over year, and it would be good to look for modest ways to use it to good ends. So an intern in the next outreachy round seems like a good “trial balloon” in this direction. Anyway, we can use this thread to track when we make any progress on discussions with Outreachy and what steps we need to take when. —Gershom

2 1

ANN: Creation of Haskell-Community list for Haskell.org Community Infrastructure Discussions
by Gershom B 09 Sep '15

09 Sep '15

Dear all, The haskell.org committee [1] had a productive week during ICFP, and at some point we'll try to write up some of the small things underway and future plans -- many things are quite tentative at the moment. However, one thing that became clear to us (well, thanks to the useful prodding of SPJ) is that we have historically participated in discussions in other venues (-cafe, reddit, etc) and then had our own internal discussions (few, seldom, and largely organizational to be honest) on the low-traffic committee [at] haskell.org mail alias. But this leaves open people wondering what those discussions are. And it also leaves open where the *designated place* to discuss haskell.org community infrastructure is. The haskell-infrastructure [2] list is very quiet and really about technical considerations. Meanwhile, -cafe, reddit and soforth are about anything and everything. So we created another list, which will be a place where we seek to have our discussions related to plans for haskell.org committee work, and where we invite everyone to join us. This new list is the haskell-community list: https://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-community The committee alias will still work, and is still the way to just write directly to the members of the committee -- no list to join. But if you wish to have a discussion about how we have things set up on haskell community infrastructure, services provided, and that we might wish to add, and how you can help (or even just what your thoughts are on how things might be done), now there is a good place for that. Hope to continue conversations with many of you there, Gershom (for the haskell.org committee) [1] https://wiki.haskell.org/Haskell.org_committee [2] http://community.galois.com/mailman/listinfo/haskell-infrastructure

2 2