[tex-live] Re: [tex-fonts] error in ec encoding

Joachim Schrod jschrod at acm.org
Thu Aug 18 20:59:57 CEST 2005

Hello Nelson,

Your idea with the checksums is great.

But, to say that different files with the same name in different
packages (that are supposedly the same also semantically) is `simply
bad archive management practice' is ridiculous. That's as if you
demand that librarians change book pages when they detect an error in
it. Life ain't as simple you try to make it. Please don't disqualify
yourself with such a blank statement, when you obviously haven't
realized the constraints of (current) CTAN management.

In particular, your rant about small disk prices is completely off the
point. Our problem is not money, our problem is that the whole CTAN
archive management (including cataloguing new entries, communicating
with uploaders, and also sysadmin work) is done by just three people. 
By the way, these three people are all on holiday now and won't answer
you this week; I'm keeping an eye on the DANTE server and on the ctan
mailing list in the mean time.

If we give you an account on a CTAN server, would _you_ (either
personally or with persons that _you_ manage) please:

 -- implement a checksumming solution
     -- storage is no problem, we've got plenty of that
     -- available is also a database (currently we use PostgreSQL), or
        CVS or Subversion or anything you need for it. We've got
        everything for you to start.
     -- Your BibTeX-headers are not a solution that will be accepted
        by submitters. Your solution must be one that DOES NOT ALTER
        SUBMITTED FILES. We are not allowed to do so on our own.
     -- Your solution MUST NOT demand headers from submitters. The `C'
        in CTAN stands for `comprehensive' -- we archive everything
     -- Some authors deny us to even add files. Therefore, your
        solution should store the meta-data somewhere else, not within
        the package.
     -- By the way, you know about the catalogue XML files and
        database, do you? That might a good starting point.

 -- implement a method that detects upload of duplicate files to CTAN
     -- Please note that this method needs heuristics, i.e., it needs
        to borrow AI methods
     -- The reason: Currently, the CTAN administration only checks for
        naming conflicts of package names. You want to add checks for
        single files that are named the same in different uploaded
        packages. Your EC.enc example says so, these are four files
        with the same name in four packages. All those packages might
        also have a file README. They can be different, can they? 
        Therefore your heuristic must be able to decide which file
        name collisions are relevant and be checked, and which
        collisions aren't.
     -- Your method MUST NOT be as simple as restricting duplicated
        file names. We have lots of names that are perfectly good in
        many packages (most of them auxiliary files like Makefile or
        README, but also some driver files, etc.)
     -- Your method must detect the owners of other versions of that
        file on CTAN and must be able to identify the reference
     -- Your method must cope with the fact that that reference
        package might not be maintained any more and that this is
        simply a spawned version.
     -- Again: Denying uploads because they don't conform to your idea
        of good archive management is a no-no because it would counter
        the whole objective of CTAN. So you have to have a process
        that makes conflicts manageable. They will exist over time.

 -- implement software support for your method above.
     -- Your method must be almost completely automated. No additional
        manual steps must be necessary for contributions. We already
        ask them a few questions; e.g., if they don't have a README
        file. E.g., your software might give us information what else
        to ask for; that would be a good help.
     -- That is important: The admin work is already too much for the
        available volunteers, we cannot add new manual work.
     -- Firing up Emacs and loading each submitted file is not
        acceptable. The software must be usable remotely from the
        command line, as this is the work environment of CTAN

 -- integrate your software support into the available software. In
    particular, in
     -- the upload software
     -- the mirroring software

 -- set up the checksumming data and the initial problems.

If you do all this, then more people will listen to your rant. Up to
then, you are simply unrealistic in your demands.

Concerning your concrete EC.enc problem,
 -- CTAN maintainers will _not_ update EC.enc in different packages.
    This is package maintenance and not archive maintenance, the
    package owners have to submit upgraded packages.
Therefore, you can help the TeX community here, as you did so often in
the past:
 -- Contact all authors of the respective packages where they got
    their EC.enc from and if they can upgrade to the current version.
 -- Of course, the authors will have to make sure that their package
    still is functional after that update.
 -- Make them upgrade their package and submit that upgrade to CTAN.
 -- Contact the TeX-Live group and ask them about the change of file
    name and content. Get the issue resolved with them. This may also
    include communication with Thomas Esser and the MikTeX folks.

With both of these actions,
 -- either getting active in CTAN management and supplying a file
    collision solution without changing the CTAN objectives,
 -- or with sorting out the EC.enc collision,
you would help a lot. And help we need and take without objections.
But rants with unrealistic demands are something different.
(If it hadn't been you, I would have simply discarded the message and
not taken the time to answer.)

Best wishes to Salt Lake City, and give my greetings to Margret,


Joachim Schrod					Email: jschrod at acm.org
Roedermark, Germany

	``How do we persuade new users that spreading fonts across the page
	like peanut butter across hot toast is not necessarily the route to
	typographic excellence?''			-- Peter Flynn

More information about the tex-live mailing list