On 08/01/11 10:28, Peter Simons wrote:
Hi Magnus,
do you think that our earlier discussion was on target, i.e. did we correctly identify the real pain points [...]?
I'm not quite sure what the problems were that we identified in earlier discussions, but I believe that the points I just explained in my previous posting are fairly accurate: our efforts aren't coordinated enough, we don't communicate enough, and our tools require way too much manual effort to get the job done.
The points you bring are the main points as I remember them, but my memory has been known to be failing in the past ;-)
[Does] our suggestions on improving our tools still make sense?
Well, I am not sure what concrete suggestions have been made. I am aware of several patches that Remy has come up with, and those patches are quite valuable because they address concrete deficiencies in cabal2arch. Still, cabal2arch is not even close to generating a working habs tree from Hackage without requiring massive human intervention. We have yet to come up with algorithms and strategies to improve that.
Those are still patches to cabal2arch, as I recall there were discussions of other complementary tools. Especially there was discussion of a tool that would allow us to get rid of our habs repo completely and instead just maintain a single file listing all packages and theirs versions. As I see it this tool could either be built on top of cabal2arch, or (thanks to the archlinux library) be written from scratch without too much difference in effort required.
I realize that my comments in this thread are only semi-helpful, because all I do is point out problems. What I should be doing is pointing out *solutions*. Unfortunately, I don't know the solutions either. Now, there are lots of smart people on this list. I was hoping that we can figure out the solutions together.
Well, we have to start somewhere. From my point of view you are basically just raising a bug. As long as you help with clarifying the issue so we can find a solution then I have no problem with your "semi-helpfulness" ;-) I think that discussing this from the point of view of what process we'd *like* to have, then see just how far away from it we are, will help us clarify what we need to do. As I've said before, the experience you have with the current tools will be helpful in this discussion. I'll see if I can dig out our old discussion to refresh my memory and see how far we got that time. If anyone beats me to searching the archives, please post the links. /M -- Magnus Therning OpenPGP: 0xAB4DFBA4 email: magnus@therning.org jabber: magnus@therning.org twitter: magthe http://therning.org/magnus