[Web Comm] Website map ...

[Auke Jilderda]

On 18 October 2006 00:57, Jason Birch wrote:
> 
> Auke, I don't want to slam CN; it does much of what it's advertised to
> do very well, and the quality of service within those bounds has been
> excellent.  However, a lot of the frustration has been around the
> ability to configure things in ways that we know are technically
> possible and are accepted standards in open source, such as (one
> example) setting up the mailing lists to reply to sender, strip
> attachments, etc.  Part of the problem is that our membership knows
> what can be done with open source tools, and is used to having the
> freedom to do it.
> 
> Jason

Understood.  And agreed, the communication part is not as flexible and
as smooth as it needs to be (see [2]).

A key question though is to what extent is the requested flexibility
critical/required and to what extent nice-to-have for OSGeo?  Properly
defining the line between worthwhile and luxury is one of the key
decisions in the infrastructure discussion.  One extreme of the spectrum
is the community doing everything themselves, allowing full control and
flexibility at the cost of volunteers managing an infrastructure instead
of working on OSGeo.  The other extreme is having a fully hosted
solution, giving limited control and flexibility while maximising the
number of people working on OSGeo.  Having read a lot of the threads and
documents, it seems OSGeo really needs something in the middle.

I know I may be (perceived to be) biased, working for a stakeholder, but
so far miss two things in the threads and discussions (but correct me if
I'm wrong, perhaps I simply overlooked it):
- It is tempting to underestimate the cost, or effort, required to run
  an infrastructure in the long run.  Setting up bits and pieces is one
  thing but keeping it up consistently at the required level of service
  under load while supporting the user base is quite another.  I wonder
  if OSGeo has talked to other large communities that maintain their own
  infrastructure about this?  If not, Brian Behlendorf (CN's CTO)
  managed the Apache Foundation's infrastructure for a long time and has
  tons of connections there.  It might be valuable to OSGeo to talk to
  Brian or somebody from the Apache Foundation to make sure the
  estimated cost, or effort, is realistic.
- Second, there are substantial features and enhancements already and
  soon available that the community does not seem to be aware of.  For
  example WebAPIs [1] to open up access to the site and increase
  flexibility, revamped discussion services [2] to unify lists and
  forums and better integrate into the user management, and CUBIT [3]
  for integrated test and builds.  These three can provide a lot of
  value to OSGeo and deserve, in my opinion, a closer look from OSGeo
  folks.  Question is who wants to see it? ;-)

Above and beyond the debate on infrastructure needs and wishes, I like
the discussion that Jody is driving, taking a much more user centric
view.  What target audiences are out there and what are their needs.
Come to think about it, that discussion should arguably pre-date the
infrastructure discussion, I think.  And it looks very good, explicitly
defining target audiences and deriving the infrastructure needs from
there.  I like the user centric view and enabling and stimulating this
user-contributor-developer evolution.


Auke

 1. Web APIs -- CN is quite aware of a managed service having limited
    flexibility and, as I recently explained to the board, CN has an
    initiative in place to open up the platform aimed solely to increase
    the flexibility of the solution.  The latest release of CEE, release
    4.5.0, is the first to start delivering this to users, introducing
    web APIs that allow users to access functionality and data via
    programming interfaces.  This latest release delivers the first set
    of APIs and subsequent releases are scheduled to add more APIs.  I'd
    love to find people who would be interested to introduce the web
    APIs and show what can and cannot be done at present, outline future
    plans, and have OSGeo tell CN how well this fits the community's
    needs and what else would be useful.

 2. Revamped discussion services -- CN Engineering has a revamped
    solution ready to be incorporated in the next major release, early
    next year.  These discussion services unify mailing lists and web
    based forums into single channels where each user can choose which
    conduit he prefers.  (I have not yet seen such a solution out there
    and think it is a great step forward; I prefer mail but other prefer
    forums and it's great to connect the two worlds.)  In addition, they
    provide a lot more flexibility in configuration and completely
    integrate into the user account management in CEE.  All in all, a
    big step forward.  I can organise a sneak preview (and a chance to
    provide feedback to the engineers developing it!) on working system
    for interested people?

 3. CUBIT -- Build and test systems as part of its managed service.  The
    way this works is that a project will have virtual build servers
    that run VMware images provided by the project, e.g. Debian
    GNU/Linux with the appropriate libraries and compilers to build the
    project for Linux and Windows with libraries and compilers to build
    the project for Windows.  Users can see a project's virtual servers,
    both which are in use and which are still free, and can reserve a
    server for their build.  Given the appropriate access rights, users
    can even reinstall virtual servers within half a minute or so.
    Naturally, you have full access to the virtual servers and can SSH
    into them.