[Desktop_architects] Cross-site bug tracking

Bryce Harrington bryce at linux-foundation.org
Tue Apr 10 10:14:46 PDT 2007


On Tue, Apr 10, 2007 at 10:23:15AM -0600, Aaron J. Seigo wrote:
> On Tuesday 10 April 2007, John Cherry wrote:
> > - Is the work on defect tracking across projects that is being
> >   done with the Launchpad site something that can be leveraged
> >   by the community?
> 
> personally, i find bugzilla and the rest of the centralized, online bug 
> systems an increasingly poor fit for Free software development. i'd love to 
> see a truly distributed bug system, perhaps bringing to bug tracking what git 
> is bringing / has brought to distributed scm.

I agree; here's something similar I proposed a while back:

http://developer.osdl.org/dev/tab/fellow_fund/fund_requests/bug_tracker.txt

> trying to bridge the various bug systems is a pragmatic approach and starts to 
> bridge the gap between various cross-stream sources, but falls far short of 
> addressing other very real issues that exist such as:

Pragmatically, getting everyone in the open source world to switch bug
trackers would be extremely difficult.  My guess is that it'd be like
what happened with the community-wide CVS->SVN migration over the past
few years; instead of everyone ending up on SVN, there came a
proliferation of VCS systems and things have gotten more diverse, not
less.  My bet is that any push to move projects from bugzilla to a
better bug tracker is going to result in a similar proliferation.  For
interoperability purposes, that could actually work counter to our goal.

This is why I tend to feel that rather than making projects change bug
trackers, it might be more productive to come up with a generic
inter-tracker communication format for bug reports, and patch in support
for that into all the major trackers.  Then no project is faced with
having to change from what they're used to - they just gain a new
feature on what they already have.

Bug trackers contain similar-ish data, so a universal bug schema ought
not be that difficult to identify.  And interlinking and trackbacks and
such sound blogging-ish; the wagon ruts are well worn there.
 
> really, it starts to look a lot like the best thing we could do is get rid of 
> the "bugs in a database" model and move to an scm based system which would 
> give us distributed tracking and possibly even distributed reporter identity 
> (a lot easier than most identity systems since there's no really "valuable" 
> information tied to the identity; it's just a convenient way to track 
> conversation). 

I'm definitely a fan of exploring non-LAMP approaches for things like
this.  Databases are handy esp. for fast complex searches or joining
data from multiple tables, but for those of us who are comfortable
mucking about in a file system, having the data as flat files (text, xml,
or whatever) has pretty cool advantages too.

However, with inotify and metadata-based search tools like Beagle, et
al, fast complex searches can be done on the filesystem directly, at
least in theory.  The bug reports themselves just become data files,
which could be browsed/edited by any number of tools - web, gui,
cmdline, etc.

In theory, databases also allow network access to the data, although no
one in their right mind would open a database for remote anonymous
read/write access.  However, technologies for doing remote file sharing,
via webdav, nfs, git/svn, email, and the like are all well known and
have rich capabilities for access control, indexing and so forth.

So yeah, I think the concept of "invent a bug report XML schema, throw
them all in a webdav filesystem, add an inotify-based search tool, and a
couple tools for adding/editing reports" would be a very interesting
approach.

> ah, if only i had more time to work on other projects. or a clone army. that'd 
> be cool. or scary, depending on your viewpoint. ;)

You said it.

Bryce




More information about the Desktop_architects mailing list