Date: Thu, 24 Jun 1999 19:01:04 -0700 (PDT) From: asami@FreeBSD.org (Satoshi - Ports Wraith - Asami) To: tg@ihf.rwth-aachen.de Cc: ports@FreeBSD.org Subject: Re: new categories: java, irc, x11-servers Message-ID: <199906250201.TAA83085@silvia.hip.berkeley.edu> In-Reply-To: <kqbte6dmbu.fsf@zabagek.ihf.rwth-aachen.de> (message from Thomas Gellekum on 24 Jun 1999 10:39:17 %2B0200) References: <199906231121.EAA51781@silvia.hip.berkeley.edu> <kqbte6dmbu.fsf@zabagek.ihf.rwth-aachen.de>
next in thread | previous in thread | raw e-mail | index | archive | help
* From: Thomas Gellekum <tg@ihf.rwth-aachen.de>
* What do you plan here? I think this is more important than creating
* new top-level categories. I'd really prefer something like, e. g.,
*
* lang/java
* lang/c
* lang/misc
* games/solitaire
* games/multiplayer
* games/amusements
* games/minesweepers
* games/misc
* net/www
* net/news
* net/irc
* net/misc
* x11/wm
* x11/servers
* x11/toolkits
* x11/misc
*
* over the current collection of directories.
*
* What would be needed apart from major surgery in the repository?
One thing to remember is that currently the biggest problem is the
size of some of the subdirectories, not the number of directories in
the toplevel. Here is the current top ten (including CVS directories
and such, so I might be off by a couple):
japanese/ 253
net/ 224
games/ 215
devel/ 169
graphics/ 137
www/ 111
misc/ 109
x11/ 108
lang/ 106
mail/ 90
Moving x11-* to under x11, www/news etc. to under net and a newly
split java into under lang will make the whole thing look a little
neater but there's not where our problem were anyway.
That said, there are several major issues here:
(1) More disagreements on what's the right categorization. Everyone
has his/her own idea of what's a good categorization, and now we
have just tossed in categorization of categories into the mix.
For instance, does mail go under net, or does it stay in the
toplevel? Do we make a new toplevel to put converters and
textproc together, or do they stay separate?
Do we put Japanese editors under japanese/editors or
editors/japanese? What do we do about Korean web browsers? Do we
go for three intermediate levels, like korean/net/www or just use
korean/www (or korean/net)? Or maybe we should collect all the
natural language support stuff under one toplevel. Then we are
going to have nls/korean/net/www for symmetry (!?) with net/www.
And maybe lang, devel etc. should really be under "computers" so
they won't be confused with natural languages and developing
nations.
Etc., etc. Basically this is not really a can of worms I'd like
to open right now. ;)
(2) Is our goal to have a tree where all leaves are at the same level,
or can they be different? (The former actually makes it much
easier to do a quick search in the entire tree -- try doing a
"find . -name Makefile | xargs grep ..." on a tree with several
large work directories.)
If it's the former, is going for another level really going to
help? Looking at the top ten list above, it will help for
japanese, games, graphics and maybe devel. But www and mail are
just going to move directly under net.
Also, if it's the former, do we go to three (or whatever) levels
all at once or start by converting some of the bigger ones first?
This could have a major impact on what's needed to change in the
framework and the time the tree has to frozen to accomplish this.
(Not to mention all the cgi scripts and stuff out there that
assume the ports tree is only two levels....)
(3) Does it really make it easier for people to find ports they want?
Ok, games/solitaire and games/minesweeper are fairly obvious, but
where do multiplayer fantasy games go if there are both
games/multiplayer and games/fastasy? Isn't what we want a better
search mechanism and a database, not more structure?
(4) And of course all the technical issues of bsd.port.*mk and others
assuming the depth of the tree. But these shouldn't be too hard
to fix. The only one I can think of offhand is some
Template/README.* files assuming all the leaves are at the same
level and there is only one level of intermediate directories in
between.
Basically, all the evidence above points to us needing a more
asymmetric tree and a better search mechanism (both for ports for
users and Makefile/PLIST type files for porters). A simple move to
three levels is not really going to solve any of the above.
-PW
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ports" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199906250201.TAA83085>