Hi everyone,
there seems to be quite a lot of room for improvement in the current Fedora LaTeX scheme. There are a lot of useful LaTeX packages that would be nice to have on Fedora, however we don't have any LaTeX package guidelines for now.
Also something should be done about the TeX Live packages. They are out of date, and some bundled packages (some of which are incomplete). IMHO the TeX compiler package should be a bare minimum with everything else shipped in their own packages, this would make updating easier.
Would anyone be interested in creating/joining a LaTeX SIG?
2009/6/11 Jussi Lehtola jussilehtola@fedoraproject.org:
Hi everyone,
there seems to be quite a lot of room for improvement in the current Fedora LaTeX scheme. There are a lot of useful LaTeX packages that would be nice to have on Fedora, however we don't have any LaTeX package guidelines for now.
Also something should be done about the TeX Live packages. They are out of date, and some bundled packages (some of which are incomplete). IMHO the TeX compiler package should be a bare minimum with everything else shipped in their own packages, this would make updating easier.
Would anyone be interested in creating/joining a LaTeX SIG?
Count me in.
Le Jeu 11 juin 2009 16:47, Jonathan Underwood a écrit :
2009/6/11 Jussi Lehtola jussilehtola@fedoraproject.org:
Would anyone be interested in creating/joining a LaTeX SIG?
Count me in.
As I wrote before, people willing to improve TEX in Fedora are welcome to use the Fonts SIG infrastructure (mailing lists, etc). The Fonts SIG is a "text-related bits" SIG in practice anyway.
There is a lot of stuff that needs to be fixed in Fedora TEX, not just the LaTeX parts.
Nicolas Mailhot wrote:
Le Jeu 11 juin 2009 16:47, Jonathan Underwood a écrit :
2009/6/11 Jussi Lehtola jussilehtola@fedoraproject.org:
Would anyone be interested in creating/joining a LaTeX SIG?
Count me in.
As I wrote before, people willing to improve TEX in Fedora are welcome to use the Fonts SIG infrastructure (mailing lists, etc). The Fonts SIG is a "text-related bits" SIG in practice anyway.
There is a lot of stuff that needs to be fixed in Fedora TEX, not just the LaTeX parts.
I'm interested. Migration to texlive was discussed here before. Died after nobody wanted to try to inspect licences of all the packages, IIRC.
2009/6/11 Neal Becker ndbecker2@gmail.com:
I'm interested. Migration to texlive was discussed here before. Died after nobody wanted to try to inspect licences of all the packages, IIRC.
Well, it's not so much that (Tom Callaway actually did go through a number of the files packaged). The problem is that the way in which texlive is put together makes it totally miserable from a distribution packaging perspective, essentially dumping a vast ammount of upstream tarballs into one tarball. This in turn makes it a really horrific job from a license auditing point of view.
In my opinion, the root of the problem(s) from a distribution packaging POV is that texlive is itself a distribution with its own package format and manager etc. which has no integration with distribution packaging systems. Roughly and broadly speaking this is how things are right now (grossly oversimplified):
1) Texlive amalgamates and consumes vast numbers of upstream packages. 2) These packages are are unpacked and built into a build tree, and texlive folk also add their own modifications and patches to integrate the whole lot into a useable system. This build tree pretty much defines the filesystem layout. 3) the buildtree componenets are distributed as texlive packages using their packaging format. For distributions the buildtree is more or less made available in a couple of big tarballs.
For any other software, Fedora would expect to be doing part 2 itself, and this is where the problem lies.
What I believe we need to work on is expanding the tool set used for 2 by texlive such that they can be used inside RPM packaging in order to allow us to control packaging ourselves, working from the true upstreams, rather than the tarballs of the texlive sandbox. The trick is to also take advantage of the integration work done by the texlive folks.
The problem is, this is an awful lot of work. I have made stabs at starting it, but, well, paid work and sleep are the enemy of Fedora contributions.
J.
"JL" == Jussi Lehtola jussilehtola@fedoraproject.org writes:
JL> Hi everyone, there seems to be quite a lot of room for improvement JL> in the current Fedora LaTeX scheme.
I will point out that there has been some discussion in this area and plans formed, but for some bizarre reason the discussion happened on the rel-eng list and not anywhere where most people would notice. Look for the thread "TeX Live 2008/9 packaging and you" at http://lists.fedoraproject.org/pipermail/rel-eng/2009-June/thread.html
What I took away from reading (but not being able to participate in) that discussion is that our texlive* packages are going to be split up and that the package review team is going to have an astonishing amount of work to do.
JL> There are a lot of useful LaTeX packages that would be nice to JL> have on Fedora, however we don't have any LaTeX package guidelines JL> for now.
Writing those is part of the proposed process.
JL> Also something should be done about the TeX Live packages.
Doing something about that is part of the proposed process.
JL> IMHO the TeX compiler package should be a bare minimum with JL> everything else shipped in their own packages, this would make JL> updating easier.
And that's what's proposed.
JL> Would anyone be interested in creating/joining a LaTeX SIG?
I don't personally care much about TeX but I have to provide a proper TeX installation to my users, so I'll be happy to help where I can.
- J<
On Thu, 2009-06-11 at 10:44 -0500, Jason L Tibbitts III wrote:
"JL" == Jussi Lehtola jussilehtola@fedoraproject.org writes:
JL> Hi everyone, there seems to be quite a lot of room for improvement JL> in the current Fedora LaTeX scheme.
I will point out that there has been some discussion in this area and plans formed, but for some bizarre reason the discussion happened on the rel-eng list and not anywhere where most people would notice. Look for the thread "TeX Live 2008/9 packaging and you" at http://lists.fedoraproject.org/pipermail/rel-eng/2009-June/thread.html
What I took away from reading (but not being able to participate in) that discussion is that our texlive* packages are going to be split up and that the package review team is going to have an astonishing amount of work to do.
On 06/02/2009 07:27 AM, Jindrich Novy wrote:
New version texlive-2008 (to be in f12):
- one single texlive package generating 3944 subpackages / 1065 MiB
Oh. My. God. Well, I'm happy that something is going on.
AFAIU from the rel-eng mailing list thread Jindrich is now working on creating the infrastructure needed to automate forming spec files for the bunch of TeX subpackages?
This amount of (sub)packages, 1/4 of the current amount in Fedora, kind of speaks for an own TeX SIG in the least :D
"JL" == Jussi Lehtola jussilehtola@fedoraproject.org writes:
JL> On 06/02/2009 07:27 AM, Jindrich Novy wrote:
New version texlive-2008 (to be in f12): one single texlive package generating 3944 subpackages / 1065 MiB
JL> Oh. My. God.
Please read the whole thread; that was the initial proposal, not the final one.
- J<
On Thu, 2009-06-11 at 11:55 -0500, Jason L Tibbitts III wrote:
"JL" == Jussi Lehtola jussilehtola@fedoraproject.org writes:
JL> On 06/02/2009 07:27 AM, Jindrich Novy wrote:
New version texlive-2008 (to be in f12): one single texlive package generating 3944 subpackages / 1065 MiB
JL> Oh. My. God.
Please read the whole thread; that was the initial proposal, not the final one.
Yes, I realize that. Maybe I should've written it more clearly :) Sorry.
It is of course clear that the single spec file route is out of the question, as it makes updating and pruning out bits and pieces a PITA.
On Thu, 2009-06-11 at 19:02 +0200, Nicolas Mailhot wrote:
Jindrich expressed fear about getting stuck in review if he split his srpm but on the fonts part at least we've been reviewing *every* *single* package submitted for review for months (usually after a few days) and in fact all the reviews that went nowhere were closed
because
the submitter went MIA, not because of lack of reviewer.
On Fedora 11 $ yum search fonts|wc -l 467
This is less than 1/4 of the amount of packages Tex Live will bring. [1] This will be a humongous project all in all. Thumbs up!
[1] http://lists.fedoraproject.org/pipermail/rel-eng/2009-June/005591.html
Le jeudi 11 juin 2009 à 20:38 +0300, Jussi Lehtola a écrit :
On Fedora 11 $ yum search fonts|wc -l 467
Actually $ repoquery --repoid=rawhide --whatprovides "font(*)" | sed "s +-[0-9]*:.*++g"|wc -l 233
And every single package listed here was re-done for F11 because we only introduced the means to autodetect font provides during this cycle (and a lot of them were re-reviewed because of srpm naming changes)
This is less than 1/4 of the amount of packages Tex Live will bring.
So what? This is not the upper limit of what can be reviewed. This upper limit is not known because we didn't manage to saturate the font review queue yet. What is missing is font packagers, not reviewers.
It is very easy for a single person to review many packages as long as they follow strictly a simple well-known pre-approved template. You basically just run diff or meld and check the licensing is acceptable.
Le jeudi 11 juin 2009 à 10:44 -0500, Jason L Tibbitts III a écrit :
"JL" == Jussi Lehtola jussilehtola@fedoraproject.org writes:
JL> Hi everyone, there seems to be quite a lot of room for improvement JL> in the current Fedora LaTeX scheme.
I will point out that there has been some discussion in this area and plans formed, but for some bizarre reason the discussion happened on the rel-eng list and not anywhere where most people would notice. Look for the thread "TeX Live 2008/9 packaging and you" at http://lists.fedoraproject.org/pipermail/rel-eng/2009-June/thread.html
What I took away from reading (but not being able to participate in) that discussion is that our texlive* packages are going to be split up and that the package review team is going to have an astonishing amount of work to do.
I didn't see this before but I can only agree with the replies: this is an insane plan. Nobody is ever going to review properly a 2.7 MiB spec file, updating will be hell, etc.
Jindrich expressed fear about getting stuck in review if he split his srpm but on the fonts part at least we've been reviewing *every* *single* package submitted for review for months (usually after a few days) and in fact all the reviews that went nowhere were closed because the submitter went MIA, not because of lack of reviewer.
Of course that can be done because we expect *simple* *standard* specs as input, which this proposal isn't.
The amount of work can probably be reduced by killing the source packages (we have srpms for that) and only shipping the most recent font formats (TEX people like to ship the same font in 2-3 different formats, that only encourages users not to convert to the latest one)
"NM" == Nicolas Mailhot nicolas.mailhot@laposte.net writes:
NM> I didn't see this before but I can only agree with the replies: NM> this is an insane plan. Nobody is ever going to review properly a NM> 2.7 MiB spec file, updating will be hell, etc.
Isn't it nice, then that the final plan is different from the initially proposed one? I see that you read the thread, so why bother commenting on something that's not currently being proposed?
- J<
Le jeudi 11 juin 2009 à 12:17 -0500, Jason L Tibbitts III a écrit :
"NM" == Nicolas Mailhot nicolas.mailhot@laposte.net writes:
NM> I didn't see this before but I can only agree with the replies: NM> this is an insane plan. Nobody is ever going to review properly a NM> 2.7 MiB spec file, updating will be hell, etc.
Isn't it nice, then that the final plan is different from the initially proposed one? I see that you read the thread, so why bother commenting on something that's not currently being proposed?
I didn't notice at first that archive threading was broken and that you had bits of the thread right and left. Also, I didn't see any clear final plan in the list archives
2009/6/11 Nicolas Mailhot nicolas.mailhot@laposte.net:
The amount of work can probably be reduced by killing the source packages (we have srpms for that) and only shipping the most recent font formats (TEX people like to ship the same font in 2-3 different formats, that only encourages users not to convert to the latest one)
The (La)TeX stack has varying levels of support for different font formats such that, at present, killing off any particular format is, sadly, probably not possible. The situation is evolving in the right direction though :).
Le jeudi 11 juin 2009 à 19:17 +0100, Jonathan Underwood a écrit :
2009/6/11 Nicolas Mailhot nicolas.mailhot@laposte.net:
The amount of work can probably be reduced by killing the source packages (we have srpms for that) and only shipping the most recent font formats (TEX people like to ship the same font in 2-3 different formats, that only encourages users not to convert to the latest one)
The (La)TeX stack has varying levels of support for different font formats such that, at present, killing off any particular format is, sadly, probably not possible. The situation is evolving in the right direction though :).
Though in Fedora we usually optimize new tech at the expense of legacy stragglers, and not the other way around :) If killing old font formats can help push a new TeX stack sooner (even if it misses a few bits that still expect legacy fonts) I'm all for it.