IRC log of fenfire on 2005-03-24

Timestamps are in UTC.

19:27:14 [fflogbot]
fflogbot has joined #fenfire
19:27:14 [niven.freenode.net]
topic is: http://fenfire.org - the channel may or may not be logged
19:27:14 [niven.freenode.net]
Users on #fenfire: fflogbot mudyc Gwl ibid antont humppake jvk benja_ majukati tuukkah
19:36:17 [benja_]
hmm
19:49:06 [benja_]
well, there is now a cron script copying the logs to eppi
19:49:21 [benja_]
every 15 minutes
20:15:14 [benja_]
http://fenfire.org/irc/fenfire/
20:32:43 [ibid]
benja_: the rekisteriseloste has broken charset, i fixed that in my copy, can you pull?
20:49:13 [ibid]
it's amazing how much clearer the screen looks after i cleaned it :)
20:49:19 [ibid]
(physically cleaned, that is)
20:59:18 [ibid]
added draft minutes
20:59:21 [ibid]
pull?
21:20:55 [benja_]
done
21:21:53 [benja_]
The requested URL /foundation/saannot.html was not found on this server.
21:29:12 [ibid]
known problem
21:29:17 [ibid]
fixing as we write
21:30:58 [ibid]
pull
21:31:10 [benja_]
ok
21:48:41 [tuukkah]
benja_, looks fine otherwise, but you use static methods all too much
21:49:08 [benja_]
tuukkah: how so?
21:49:17 [ibid]
tuukkah: as you can see, i preempted you. sorry :)
21:49:37 [ibid]
benja had better response time :)
21:50:10 [tuukkah]
ibid, I was still waiting for your amended patch in the morning
21:50:21 [ibid]
yeah, i fell asleep
21:50:39 [ibid]
today afternoon, it seemed like benja was the faster route :)
21:50:39 [tuukkah]
benja_, couldn't rolecontext be an object?
21:51:04 [benja_]
tuukkah: that would not be in line with how other javolution contexts work
21:51:24 [benja_]
(they're objects, but using them happens through static methods)
21:51:51 [tuukkah]
benja_, ok :-)
21:52:00 [tuukkah]
ibid, I don't know what you're talking about
21:52:10 [ibid]
benja was talking :)
21:52:11 [benja_]
I really don't want to break convention there
21:52:39 [benja_]
tuukkah: I applied ibid's changes to fenfire.org
21:52:47 [tuukkah]
it's not that bad if the class is forwarding the messages to some object
21:52:47 [benja_]
(that's what ibid was talking about)
21:53:25 [tuukkah]
I don't have a problem with, I don't like applying patches or fixing charsets
21:53:32 [ibid]
:)
21:53:46 [ibid]
benja_ talked about adding me to the group
21:53:52 [ibid]
tuukkah: have you looked at the pages?
21:53:59 [benja_]
yeah, I see no reason not to
21:54:00 [ibid]
(the unix group, that is)
21:54:06 [benja_]
someone needs to ping the admins
21:54:15 [ibid]
somebody who's already in it
21:54:21 [benja_]
(if there isn't some userspace tool that group members can use that I'm not aware about)
21:54:46 [ibid]
there is a userspace tool but only root can use it :)
21:57:11 [ibid]
the asterisks do look a little amusing, though
22:07:12 [ibid]
pull, please (fixed the displays)
22:08:27 [benja_]
'kay
22:09:50 [ibid]
(reminding benja_ on doing the joining bit:)
22:15:09 [benja_]
should we have a URI identifying the Fenfire project?
22:15:50 [tuukkah]
http://fenfire.org?
22:16:23 [tuukkah]
oh, that identifies a web page location
22:16:27 [benja_]
tuukkah: which URI would identify the homepage, then
22:16:33 [ibid]
benja_: for what purpose? (genuinely asking:)
22:16:40 [tuukkah]
I'm close to failing the course ;-)
22:16:49 [benja_]
tuukkah: =)
22:17:02 [benja_]
ibid: I'm writing a DOAP file (Description of a Project, like FOAF) for Fenfire
22:17:23 [benja_]
if we pick a canonical URI, that makes the project somewhat easier to refer to in RDF graphs
22:17:26 [tuukkah]
DOAP is famous from planet debian :-)
22:17:36 [benja_]
doap:homepage is an inverse functional property
22:17:46 [benja_]
i.e., used to uniquely identify the project
22:18:01 [benja_]
and everybody knows that <http://fenfire.org/> is our homepage
22:18:32 [tuukkah]
same trick as foaf:mbox? "the first project that had fenfire.org as their homepage"
22:18:42 [benja_]
y
22:20:43 [benja_]
one way is to use http://fenfire.org/#something
22:22:14 [benja_]
say, http://fenfire.org/#project
22:22:34 [ibid]
where in planet debian is a doap?
22:22:45 [benja_]
another way is, of course, to use a bnode
22:37:58 [benja_]
http://fenfire.org/doap.turtle
22:58:23 [benja_]
RoleContext implemented :-)
23:18:03 [tuukkah]
ibid, the developer of doap blogs there
23:18:21 [ibid]
who?
23:18:47 [benja_]
edd dumbill?
23:18:50 [tuukkah]
yes
23:18:52 [ibid]
oh
23:18:54 [ibid]
ok
23:21:42 [benja_]
I really liked earlier today to cd to the darcs archive on the web server and say, pull from this other http uri
23:22:21 [benja_]
in that kind of scenario, darcs feels like a versioning extension for http =]
23:22:32 [ibid]
:)
23:22:40 [ibid]
oh, now i get your exclamation :)
23:22:47 [benja_]
*g* :)
23:23:04 [benja_]
I think this should be an inspiration to Storm
23:23:38 [ibid]
* ibid has been pondering doing a nih-system based on some ideas in storm
23:24:06 [benja_]
hm, how do you do a system invented somewhere else? ;-)
23:24:31 [benja_]
but, do elaborate
23:25:12 [ibid]
nih in the same sense that i ponder doing my own web browser
23:25:34 [ibid]
ie., i'd rather start afresh than work on a previous design
23:25:46 [ibid]
i think it has to do with my 'world-building' attitude
23:26:14 [ibid]
but something fairly standalone anyway :)
23:26:24 [benja_]
things I'm missing in darcs (and I'm aware they're out-of-scope for the source code revision control system darcs is) are, not having rigid repository boundaries; having URIs for patches and versions; having rdf to track when you branch from somewhere else -- i.e., "what are places that branch from here?"
23:27:04 [ibid]
what does 'rigid repository boundaries' mean?
23:28:11 [benja_]
in darcs, it's clear where in the hierarchy a repository starts (the dir containing the _darcs dir) and where it ends (subdirs containing a _darcs dir are not part of repo) (and also non-added dirs and files aren't)
23:28:42 [benja_]
I'd like not to have the concept of a repo, if possible, only dirs and files with versions
23:28:51 [benja_]
hmm, this is a bit fuzzy
23:29:02 [benja_]
not clear how it interacts with the need for atomic commits
23:29:15 [benja_]
(i.e., changes to a number of files/dirs should be a single commit)
23:29:51 [ibid]
ah, you wanted to get rid of the boundaries. i wondered what kind of boundaries could you mean if the current ones are not enough ;)
23:30:07 [benja_]
:)
23:30:17 [ibid]
benja_: essentially, you want a versioned file system with acid transactions?
23:30:21 [ibid]
benja_: ie. a database :)
23:31:01 [benja_]
no :)
23:31:18 [ibid]
well, that's what it sounds like
23:31:45 [ibid]
(not a *relational* database, true:)
23:32:01 [benja_]
well, I think perhaps you misunderstand what I mean about atomic commits -- darcs has them, cvs doesn't, we need to keep them
23:32:18 [ibid]
atomic commits are what acid transactions are about
23:32:18 [benja_]
I want something that's less like a file system than darcs is, I think =)
23:33:02 [ibid]
acid = atomic, consistent, isolated, durable
23:33:21 [ibid]
in case you haven't heard, which you probably have :)
23:33:56 [benja_]
:)
23:34:10 [ibid]
have you?
23:34:14 [benja_]
yes
23:34:22 [ibid]
or should i lecture... probably not, then :)
23:34:41 [benja_]
it has even been part of the content of the distributed systems course I've recently taken
23:34:46 [benja_]
ibid: some other time :)
23:34:49 [ibid]
:)
23:35:06 [benja_]
I'm more interested in discussion about darcs and beyond...
23:35:13 [ibid]
* ibid is hot on transactions currently, thanks to spj et co's paper on software transactional memory
23:35:21 [benja_]
I want to be able to branch a single web page, for example
23:35:54 [ibid]
ok, fine-grained, nestable repos? :)
23:36:00 [benja_]
hmm
23:36:26 [benja_]
I mean, the single web page wouldn't be considered its own repo by its author
23:36:42 [ibid]
i got that
23:36:54 [ibid]
it's just the way my math-trained mind works :)
23:36:56 [benja_]
more interesting case: consider http://himalia.it.jyu.fi/darcs/fenfire/metacode/send_darcs_diffs.py :)
23:37:04 [benja_]
mhm
23:37:11 [ibid]
the author has a repo, then anybody can consider any "subset" of the repo a repo
23:37:19 [benja_]
yeah, that's what I mean
23:37:36 [ibid]
with proper subrepo semantics, of course
23:37:40 [benja_]
(cvs is somewhat like that actually)
23:37:46 [benja_]
ibid: subrepo semantics?
23:38:16 [ibid]
stuff like if i commit to the single-file repository, it is visible as a single-file commit in all superrepos
23:38:29 [benja_]
yeah
23:38:32 [ibid]
self-evident stuff, really :)
23:38:41 [ibid]
but that's how my math-trained mind works ;)
23:39:16 [ibid]
actually, a file portion repo would be nice, too
23:39:22 [ibid]
but becomes nasty
23:39:29 [benja_]
that's how your math-trained mind works ;)
23:39:37 [ibid]
at least with the unixish definition of file
23:39:40 [ibid]
benja_: more or less :)
23:39:43 [benja_]
=)
23:39:58 [benja_]
an interesting question is versioning of items in an rdf graph
23:40:04 [benja_]
(interesting and relevant to us)
23:40:31 [benja_]
basically, you must be able to take an arbitrary function selecting part of a graph, and then consider the history of the so-selected subgraph
23:40:47 [benja_]
since there are no rigid boundaries in hyperstructure, which is the point =)
23:41:05 [ibid]
so versioning should be subspace-invariant
23:41:19 [ibid]
or composable, or ...
23:41:20 [benja_]
what does invariant mean?
23:41:29 [ibid]
literally, not changing
23:41:56 [ibid]
as a technical term: a X-invariant is a property that stays the same even if the pov-X changes
23:42:09 [benja_]
ibid: what does it mean *here*?
23:42:24 [ibid]
something good :)
23:42:29 [benja_]
lol
23:42:48 [ibid]
it's rather fuzzy in my head currently
23:42:53 [benja_]
ok...
23:43:11 [ibid]
to describe it properly, i'd really have to solve the problem
23:43:25 [ibid]
since defining subspace-invariance is probably a large part of the solution :)
23:43:36 [ibid]
at least in theoretical terms, dunno about implementation
23:43:43 [benja_]
another issue is the relationship between patches and versions
23:43:59 [benja_]
I really want to be able to browse the "past versions of a page"
23:44:23 [ibid]
and link to a specific version, as well as "the current" version
23:44:31 [benja_]
true, yes
23:45:06 [ibid]
i suspect that under the hyperstructure versioning problem there lay several papers on graph theory :)
23:45:17 [benja_]
but it's also useful to have patches as a first-class concept
23:45:34 [ibid]
a patch is a node?
23:46:02 [benja_]
?
23:46:23 [benja_]
I was thinking about darcs being patch-centric rather than version-centric
23:46:37 [benja_]
although a particular realization of a patch always has a version as context
23:46:41 [ibid]
i take it that first-class patch means that a patch is represented inside the versioned space
23:46:50 [benja_]
no
23:46:52 [benja_]
sorry
23:46:58 [ibid]
that's at least how my pl-trained brain works :)
23:47:28 [benja_]
in cvs, you have versions, and implicitly have patches as the gaps between the versions
23:47:49 [benja_]
in darcs, you have patches, and implicitly the versions as the gaps between the patches ;-)
23:48:02 [benja_]
you want to be able to link to both versions and patches, probably
23:48:42 [benja_]
"see this version"; "please apply this patch"; "this patch broke it because..."
23:53:00 [benja_]
"these patches have been applied"; "this patch has not been applied yet"
23:53:41 [ibid]
okay, if patches are first-class then versions are second-class :)
23:53:51 [benja_]
why?
23:54:05 [benja_]
why can't both be first-class?
23:54:59 [ibid]
then either patches are versions, versions are patches, or both are instances of some common superconcept
23:55:11 [ibid]
i don't say it cannot be
23:56:17 [benja_]
ibid: I think they're disjoint, and while they're related, they're not instances of the same concept
23:57:09 [ibid]
i mean, "first class X" implies, at least in pk parlance, a some sense of subsumption between all things first class
23:57:30 [ibid]
like, if functions are first class values, then they can be treated like any other value
23:57:40 [ibid]
s/pk/pl/
23:58:12 [ibid]
well, it doesn't matter
23:58:35 [ibid]
i think i need to go to sleep :)
23:58:36 [benja_]
they should both have URIs :-)
23:58:40 [benja_]
ok :-/