> i think that we have several problems here which to some extend
> pulling the ship into different directions; i think we first need to
> dis-tangle some of this a bit before it really makes sense to talk
> models (or even approaches)
I won't quote most of his post but I agree wholeheartedly with his
> - one of the major driving forces of latex is not necessarily the
> quality of its output but the exchangeability and portability of its
> input while maintaining a reasonable quality output and (with the
> necessary skill/knowledge (luck?)) being able to produce without too
> much fuss really high quality output from a nearly unchanged source
> - i claim that we more or less actually achieved this, it is again
Yes, and praise where praise is due! Good job!
> - and neither has be front matter stuff been addressed unfortunately
Well, we're getting to that.
> for this block of the core we provide very strict consistency checks
> and documents that used this stuff two years ago will run now and will
> do so in with the december release (will produce identical output) ---
I think this is essential.
> this stability is extremely important in the latex community, even
> upward compatible changes are considered bad by the majority of the
> in my opinion the "core" whatever this is at a certain point in time
> can only be maintained by a cathedral solution, and staying with the
> picture, without much better arguments as those brought forward so
> far, i wouldn't give my blessing to anything else.
> second, the linux example is a red herring. its development and
> implementation model works because it doesn't effect peoples use of
> the system as it is the single system that is effected if you upgrade
> your linux (but your ip protocol stays the same it is still tcpic if
> not the linux model would die too); with latex the situation is that
> the "system" is not the single latex installation but the combination
> of all those latex systems out there that you like to communicate
I agree. The model isn't applicable everywhere.
> i do see some good reasons for extending the "core" distribution but
> it would require new maintenance and support models and it would need
> somewhat conservative people that understand why it can be dangerous
> to fix the spacing in article.cls (even though we all know that
> design-wise it is mostly rubbish)
Yes. It's good to mention that conservative people are needed!
> - stability within the distribution, ie all packages would need to
> work with each other in any legal combination
Yes. This is a problem with some of the contrib stuff---not all
combinations work together OK, sometimes the order in which they are
loaded is important etc.
> - clearly defined and supported maintenance cycles such as the core
> LaTeX nowadays has (they may be longer, eg a year, but one should
> not go back to the old days of 2.09 maintenance which was applied
> whenever it was raining in California)
Perhaps less effort could be put into rapid updates (twice a year) and
use the save labour to implement some of the improvements we've been
talking about---isn't once a year enough?
A problem I have with updating LaTeX is deciding whether to make the
(substantial, if one includes non-core stuff) effort of updating
EVERYTHING (installing from scratch is easy enough, even trivial, at
least on VMS with Ralf G"artner's distribution, but I want to keep
customisations) OR finding the 1% or whatever which has actually changed
and replacing that, which all told is about the same amount of work.
There is also the issue of updates which can be made simply by replacing
files and those which require a rebuild in some sense.
Again, updating the core stuff automatically is easy enough, but we want
Perhaps one could move to the following update model.
Once a year there is a new version, as there is now twice a year.
Whenever bugs are fixed which can be fixed simply by replacing files,
these are announced. There can be a patch web page, a mailing list,
whatever. Users can decide whether they need to download the new file
(or edit their current one by hand, perhaps, if it's a trivial change)
or wait until they next do a `real' upgrade, which will of course
include all of these fix-by-replacing-this-file changes. When the
`real' upgrade arrives, it will also indicate to what extent it
o is just a set of replacement files
o includes some new files
o requires a rebuild
If no rebuild is required, people who like a rapid update could have it
even more rapidly than at present. Once a year, one could still do
everything in one go, which is probably often enough for those who
prefer everything in one go anyway.
Perhaps some folks will note a similarity to the VMS update model: there
are OS upgrades periodically, but patches are announced when they are
available. There is a patch mailing list and they can be downloaded via
the net. Each maintainer can decide if the patch is important enough to
be installed now etc. Periodically there are maintenance releases (like
7.1-2 which is coming in January or so) which are essentially
collections of all patches since the last release (but also perhaps
provide minor changes like additional hardware support) while major OS
upgrades (like 7.2 next spring) are less frequent, introduce new
features and require a `rebuild' in some sense.
This seems to work well. I guess there are about the same number of VMS
users as LaTeX users(?) and about the same mixture of folks who are
doing essentially the same stuff they were ten years ago and are
interested in keeping it going without change under the new version and
innovators who make use of many new features when they come out.
As a user and lover of VMS and LaTeX, I notice that I feel more
comfortable with keeping my OS upgraded as than I do keeping [TEXMF...]
upgraded, although of course the former is actually much more complex
(especially when one includes layered products like compilers etc and
considers the interaction between various bits). Of course, VMS is
designed quite differently than, for example, the unix `bag of tools'
philosophy. I guess I want less a bag of tools, a bazaar, and more
something like the cathedral, which seems to work well with VMS. Since
LaTeX is essentially quite restricted in its aims, I think it's worth
going for a cathedral model and bringing as much stuff as possible under
Phillip Helbig Email ......... [log in to unmask]
Nuffield Radio Astronomy Laboratories Tel. .... +44 1477 571 321 (ext. 297)
Jodrell Bank Fax ................ +44 1477 571 618
Macclesfield Telex ................ 36149 JODREL G
UK-Cheshire SK11 9DL Web ... http://www.jb.man.ac.uk/~pjh/
My opinions are not necessarily those of NRAL or the University of Manchester.