[needs-packaging] Texlive installer package should be created

Status changed to 'Confirmed' because the bug affects multiple users.

*** This is an automated message ***

This bug is tagged needs-packaging which identifies it as a request for a new package in Ubuntu. As a part of the managing needs-packaging bug reports specification, https://wiki.ubuntu.com/QATeam/Specs/NeedsPackagingBugs, all needs-packaging bug reports have Wishlist importance. Subsequently, I'm setting this bug's status to Wishlist.

As of 2012-03-27 there is an update in Debian for TexLive (2011 version), it is quite possible that it even makes it in time for Ubuntu 12.04 (see bug#712521) .

So in a sense, it feels like this bug is no longer valid.

I see that the main point here is to prevent the same thing from hapenning again, namely having TeXLive lag 2 years or more behind the upstream version.

But still I feel that stability is more important in official packages than anything else, and I can think of an extra reason for TeXLive; it has many bugs of its own to worry about.

So if the same line of though was carried to other similar cases, simply by virtue of easing a laborious task (like downloading texlive installer and running it from command line), it would give a general feeling to end users that officially supported packages are not stable.

An appoach like the one proposed here would simply add nothing in terms of stability. There are unoficially supported ppas for things like the one you mention (updated versions of programs, etc).

I still see the needs-packaging as valid, following the reasoning below:

1) Most likely, TL2011 will not make it in precise. So, we gonna be stuck with 2009 until october. By that time we will get 2011... but texlive 2012 is goning to be out by june. Not a single linux distro has ever ever succeeded shipping an up to date texlive... Hence, saying that the bug is no longer valid does not feel very convincing.

2) As you mention, the issue is not TL2011, is remaining stuck with TL2011 until 2017... I do not blame the packagers at all... the problem is that TL is a huge thing. It is for a reason that they have incorporated its own package management into it in the end.

3) Is the debian TL really so much better than the original in terms of stability? Truly debian has a big patchset against texlive, but many things are indeed trivial (e.g. using the paper lib). The other way round older TL editions have many broken packages, which get progressively fixed. Staying with older TL, means that one ends up having tons of sty files scattered around to get desired behaviors. Furthermore, TL may be buggy, but is rather reactive at shipping the updated stuff when something broken is found.

4) Older TexLive means that no-one in debian/ubuntu is ever getting to testing and helping the progress of luatex

5) Is there any statistical information about people using TeX? TeX is already a rather techincal thing. So the percentage of people using TeX on ubuntu is already a small fraction of all ubuntu users and the fraction with higher programming skills. Can it be the case that a large fraction of this small fraction of users do not install the debianized texlive but rely on the original texlive to have the up-to-date thing?

6) Most important: and if you agree on this, please change the bug title accordingly. IMHO the biggest problem with the current TL packaging is package names and dependencies. Applications that should depend on some tex features (e.g. availability of latex, etc), end up depending on some specifix texlive debian package. This makes it quite difficult to set up ppa packages with an alternative tex distribution or an alternative packaging of texlive. Possibly, virtual packages should be used to provide tex features.

@Sergio

I mostly agree with your comments, and I would be the first to try out a PPA on the likes of your proposal (I am quite lazy myself :). But I still feel that a PPA like that does not belong to the main repository, just like in so many other cases. Do you see a new official ppa for every new version of say LibreOffice, which has admittedly a significantly larger audience?

For a PPA to make it into the main repositories, it needs to be well tested to make sure it works not just in the developers' computers, but in the large and diverse user base. Just wrapping around an installer and creating a ppa doesn't help much, not unless the packages are tested... That on its own is a huge undertaking (~3GB of packages), and is the min reason TeXLive lags behind to begin with.

Anyway, considering your comment #5, indeed I would agree TeXLive users should be statistically biased towards the technically skilled side (most linux users really), and installing an up-to-date version through the TeXLive installer shouldn't be that much of a problem.

I for one have tried it twice or thrice, and unfortunately it never worked... I always had issues with kpathsea, so effectively I couldn't even compile trivially simple .tex files... And I would spend hours searching the internet for solutions, trying things in the command line, messing up my fonts cache. By that time I would have greatly passed the point of even considering manual installation a problem.

You see, that is my point, the installer might work for you and most other users, but TeXLive is really really hard to test and make sure it is going to work (even the fundamentals) in virtually all users. And if it doesn't, if installing it manually feels difficult then fixing the problem should feel impossible.

If however you still insist on your initial view, let me know so I can revert the status change of your bug report.

I have noticed that fedora has decided to give up the traditional 'monolithic' packaging of texlive and take a way that looks rather interesting, namely the semi-automatized conversion of packages from the texlive package manager to rpm.

See http://fedoraproject.org/wiki/Features/TeXLive
"Upgrade TeX Live to at least 2011. Use packaging based on upstream metadata and don't use monolithic build any more."

This is not yet ready for the large public, but already alive, working and testable. Should be in the beefy miracle.

Seems a rather interesting approach. If it works for them, It will make redhat forget for ever the times when their tex was always 2-4 years late. Also it gives an amazing granularity on what gets installed: you can install schemes, but also go down to things like
# yum install 'tex(epsfig.sty)'
to install a single latex style

I now wonder if their infrastructure can be converted to package to deb.

Please give advice about making a blueprint with Fedora's approach.

Just noticed that with this new scheme, they have a testing repo for texlive 2012 already. Which means that they will be able to ship TL2012 at the same time TL ships it.