Fw: Re: [OSGeo-Discuss] Proposal to find an alternative to Collabnet

[Jo Walsh]

I'm not sure why this response of Brian's didn't get approved yet -
am forwarding it for completeness sake - and I'm writing down my
version of 'the WebCom perspective' right now ... :)

----- Forwarded message from Brian Behlendorf <brian at collab.net> -----

Date: Sat, 15 Jul 2006 15:25:24 -0700 (PDT)
From: Brian Behlendorf <brian at collab.net>
To: Jo Walsh <jo at frot.org>
Cc: discuss at mail.osgeo.org, akelly at collab.net
Subject: Re: [OSGeo-Discuss] Proposal to find an alternative to Collabnet

I appreciate being cc'd on the email, Jo; I'll reply, though I'll admit 
that I can not claim to be writing this in the awareness of every other 
message stated in this thread or on the topic.  With that said:

On Sat, 15 Jul 2006, Jo Walsh wrote:
>dear Cameron, all,
>On Sun, Jul 16, 2006 at 04:49:49AM +1000, Cameron Shorter wrote:
>>Ethically, I'd prefer to see Open Source projects use Open Source tools.
>
>Amen, brother! Open Source ideology dictates that application code
>wants to be free. Companies create "added value" by providing managed
>services built on free software. Clients stick around because the
>business relationship improves their lives, not because they're
>"locked in". So I've never understood why the CollabNet codebase is
>not managed on an open source basis; perhaps it would be more
>conducive to client loyalty (given CN's clients are software developers
>with a itch to fix their own problems!) if it were.

I'm not going to make any friends here by repeating the reasons why 
companies don't always release all their source code.  The number of 
profitable companies who do, though, are few, and even they will hold on 
to something defensible.  You might ask why Red Hat or JBoss won't open 
source their training materials, for example.  And I know of no single 
Application Service Provider like ours who provides their entire stack of 
software as open source.

And how many of you develop on Windows?  Or Macs?

So today, there's a line - we invest millions of dollars in salaried time 
per year in Subversion, given away under a BSD license; and above that, we 
charge like anyone else would.  That line may change over time.  I hope it 
does, personally, and continue to work on the justification for it.

I would presume that what any open source community cares most about is 
keeping the community healthy and able to create high quality code, and 
have fun doing it.  If tools like CollabNet's get in the way of that, 
shame on us and pursuing alternatives makes a lot of sense.  If not, 
though, is ideology enough of a reason to make a change like this?

Realize that from its inception in 1995 to about 2002 I volunteered as the 
primary sysadmin for Apache.org.  There was an infrastructure team who 
helped out, but everything from the hardware to OS to the mail systems and 
lists and CVS and IZ installation were my responsibility.  While there was 
a lot of thrill to being so deep in the machinery, I dreaded every 
vacation or business trip (if something happened while I was away...), had 
nightmares about crashing disks, and spent lots of late nights and 
weekends dealing with spam attacks or web robots gone wild.  And the hack 
attempts, too.  Fun.  I burned out on it, big time.  There's now a healthy 
number of Apache members who continue to volunteer sysadmin, but they have 
a burn-out half-life of about 6 months.  In retrospect, I wasn't nearly 
the coder that Dean Gaudet or Justin Erenkrantz etc. are, so volunteering 
to sysadmin was a very satisfying way for me to contribute; but the people 
today who are very active sysadmin volunteers are those who aren't 
otherwise spending their time on writing free software.

Not every project is going to be like Apache (Apache's unique needs are 
one reason I never even suggested they be on CollabNet).  But there is 
enough about these needs that are common, that I felt like I really wanted 
to solve that problem for the open source community.  Not being funded by 
academia or independently wealthy, the only path I could see to doing that 
was to start a company to do it.  With all the compromises that entails.

It always feels wrong to me to tell people they shouldn't pursue something 
they want to pursue, that they have a passion for.  If you have a passion 
for building and supporting the plumbing for community, go for it.  You 
could even set it up today, in parallel to the existing CollabNet 
infrastructure, to get your feet wet with it and then make an informed 
decision about switching.

On to specific points you raise:

>- There's a great deal of interlinking / mutual dependence / mutual
> redistribution between different projects - this is a byproduct of
> being a 'stack' - where a prospective OSGeo user needs
> Geotools+Geoserver+Mapbuilder, or GDAL+Mapserver+Mapbender, or
> GDAL+OSSIM+osgPlanet, to build a whole application. What we've seen
> of CollabNet's approach keeps projects in neat rigid sandboxes, and
> that's not good for us - there's always been a lot of cross-project
> developer energy and commitment - the foundation *arises* from that
> and the tools should be reflecting it, not trying to engineer it
> into existence (or failing to...)

Huh.  One of the things I thought we do pretty well with CEE was the 
ability to build cross-project communities, through the use of parent 
projects, project categories, domain-wide groups, and other ways to group 
things across projects.  You put things in projects to give them a sense 
of shape; but nothing says that one project's build process can't build 
the code managed at another.  It's all hyperlinks.  Sounds like you want 
to create composite projects, where the source code, discussions, 
issues/defects, etc., in that composite project are all about the 
composite, but if you find a defect in a particular component in that 
project, you report it against the component's project.  If, on the other 
hand, you want them all in one project, that works too.  OpenOffice is 
still largely like that.  It just seems to me that community modularity is 
as important as code modularity.  Maybe there's something CollabNet can do 
differently on osgeo.org to improve this?

>- There's a wider focus than on software - in educational resources
> and geodata packaging and distribution which are indispensable to
> getting open source geospatial tools into real peoples' hands. CN
> doesn't provide tools for collaborative documentation management, or
> processing / bandwidth for serious data munging and serving - so
> OSGeo is obliged to look elsewhere for tools and hosting to do this.

Well, sure.  Our focus is on software development; not on arbitrary 
services.  I think we do content management alright, but not near what 
you'd find even in open source CM packages.  But, those tools aren't 
focused on software development either.  Running these in parallel is fine 
with us, many customers do; and in future releases we'll be providing web 
services APIs so you can synchronize logins and permissions, too.

	Brian

----- End forwarded message -----

-- 
ghug is my email archiving bot. if you see it cc'd on this email,
please leave it cc'd, that will help me a lot. http://frot.org/ghug