No subject

Sat Nov 26 22:11:05 UTC 2011

not as strong as you might find in alternative tool suites. Take the Atlass=
ian suite for example. With JIRA, svn, FishEye, Crucible and Confluence, yo=
u have a fully integrated and coherent system that covers source code repos=
itory, issue tracking, code review and browsing, collaborative document/con=
tent wiki. If you add Bamboo, you even get continuous integration support. =
There's other stuff too. Now, I don't want to look like I'm pit=
ching for a move to Atlassian, but having used their tools in anger at my c=
urrent employer, I can only say that they seem like a very good fit for LSB=
's needs. My understanding is that that Atlassian also state that they =
potentially offer free use for open source projects and I'd be surprise=
d if they didn't extend this to the LSB. They even have a fully hosted =
solution, something that might be seen as a positive in light of the events=
 of the past year or so. With offerings like these available, I find myself=
 looking at what LSB is currently using and I can't ignore the feeling =
it gives me, that LSB is not keeping up with modern tools, etc. I can fully=
 appreciate all the practical reasons why things are as they are and the am=
ount of work involved in making a switch. My main point here is that things=
 like this leave a less than stellar impression with people and it may be a=
 contributor to the perception of LSB not being relevant or up to date. It =
is certainly something that bugs me whenever I have to interact with the ex=
isting tools (I don't even bother with bzr - got enough new tools to le=
arn already).<br>

I&#39;m aware the above may evoke some fairly strong reactions in some. I&#=
39;m more interested in discussions about what realistic options might be a=
vailable and how to move LSB forward to make it more successful. Any toes t=
rodden on were entirely unintentional. ;)</blockquote>

<div><br></div><div><br></div><div>Well, I can give you one strong reaction=
: I&#39;ve been forced to switch to Jira &quot;at work&quot; and utterly de=
test it. =A0This is issue tracking for managers - it&#39;s unpleasant to ac=
tually track bugs for developers, but lots of lovely reports and metrics th=
at can be embedded in the detestable powerpoint slides that at least 2/3 of=
 the companies in the world seem addicted to. =A0There... that out of the w=
ay :) =A0(I am playing that for effect, of course)</div>

<div><br></div><div>I went in assuming switching to an integrated suite isn=
&#39;t happening. =A0 Quite simply, the choice of bzr holds us back. =A0I t=
ry really really hard to convince myself bzr is decent. I <i>like </i>it, b=
ut that seems to be purely an emotional reaction. =A0In terms of integratio=
n with other tools, except in a subservient way where bzr can read git, hg =
and svn repositories pretty well, it&#39;s poor; performance is terrible; a=
nd LF, in restoring their infrastructure, has chosen to deploy it in the ab=
solutely most inefficient way so it&#39;s now beyond terrible. I&#39;ve ask=
ed for the bzr:// protocol to be implemented to help a little, deployment i=
s trivial (either launch bzr in server mode once or do it on-demand out of =
whatever-passes-for-inetd-these days, and open a hole for the one port in t=
he firewall), but there&#39;s been no response. The only place bzr integrat=
es well with other tools is Canonical&#39;s Launchpad, and that&#39;s not r=
eally available to use unless we choose to host everything there - Launchpa=
d is open source, but in the sense of &quot;source is available for you to =
hack on&quot;, not in the sense of &quot;you should deploy your own Launchp=
ad-like site&quot;. =A0 The code review tool I popped up a demo of, in the =
hopes of actually providing /something/, doesn&#39;t have much support for =
bzr, because nobody really seems to care (code review? why don&#39;t you ju=
st use Launchpad?). =A0And performance when combined with bzr is so bad we =
should probably just throw it away and not even play with it, to be brutal.=
 =A0Gitorious? =A0Not if bzr is the version control. =A0Almost anything els=
e? =A0Not if bzr is the version control, not if you want it to work well. =
=A0Usually things get a shim contributed by a bzr enthusiast that implement=
s the basic pieces of the story, but misses advanced uses. =A0</div>

<div><br></div><div>By the way, we do have a sort-of-CI scheme... based on =
buildbot, it runs scheduled builds plus on-demand whenever there are checki=
ns, and is also used to request release-targeted builds. =A0It&#39;s a pain=
 to understand the setup, but as a contributor you don&#39;t need to, just =
commit and it builds.</div>

<div><br></div><div>So without being offended at all, this was all I could =
think of without any real support or any real chance at serious integration=
. =A0And frankly, there&#39;s so little engagement at the moment it&#39;s n=
ot worth even talking about forklifting the infrastructure onto something n=
ew; even restoring what we had has now passed five months without being com=
plete (which says quite a lot by itself I think).</div>



More information about the lsb-discuss mailing list