Theo Van Dinter writes:
> On Mon, Apr 17, 2006 at 12:31:17AM -0400, Warren Togami wrote:
> > Problems
> > ========
> > 1) It is only really usable on Linux/Unix hosts, or maybe Cygwin with a 
> > bunch of effort, although not easily with any Windows clients.  Even if 
> > you are a Linux user, it is *TOO HARD* to setup and get it to work properly.
> 
> I may be a bit biased, but it takes all of 5 minutes to get setup,
> though our documentation may not make it look that simple.  The scripts
> to do the heavy lifting are available in the source tree.

Hmm.  Given that about 75% of the submitters felt the need to reinvent
their own scripts, I suspect it is not quite as simple as all that ;)

> > 2) As a result, most if not all people sorting corpora and participating 
> > in nightly mass checks and corpus scoring are somehow related to a 
> > single demographic: computer hackers.
> 
> True, though I don't think that's a result of #1.  People find building a
> corpus difficult.  Maintaining that corpus properly is even more difficult
> IMO (virus/worm mails, bounces, misfiled messages, etc).  Even those of
> us who are currently involved with mass-checks find this problematic.

That's definitely something that can be taken care of at the other end --
in the worst case, we just need to not use submissions from people who
display an inability to take care of their corpora.

There's a wiki page describing quite a few tricks to detect such bad
mass-check logs.

> > 1) MUST be able to run on Windows, where most normal users are.
> [...]
> > Possible Implementation Details
> > ===============================
> > 1) Implementing this really wouldn't be all that hard because you would 
> > use existing components like perl, spamassassin, ssh, and rsync.
> 
> ssh isn't required, but perl, the SA required perl modules, and rsync are.

> [...]
> > 5) The use of existing FOSS components and a cross-platform toolkit like 
> > gtk+ or qt would allow this to build on Linux too.
> 
> 6) No one in the current development group has experience doing this
> kind of stuff (as far as I'm aware), so there will likely be no one
> available to maintain it.

and who's to say the SoC student might not decide to carry on working on
the project? ;)  see http://tirania.org/blog/archive/2006/Apr-13.html :

  'Marek completed the required tasks for an MSBuild  implementation, he
  went back to school but came back to the project and leads the effort to
  complement the work that he did over the summer.

  Marek is now working with other Mono developers to improve his
  implementation. He is a regular in the development channels. '


> 7) I don't know if the SA project could make such a package available.
> It requires non-ASL licensed code (perl, perl modules, rsync, etc,)
> which may not be compatible.

Now that *is* a good point, we have quite a solid restriction on
what we can distribute :(


All the same, I think the general idea is a good one.  

I was thinking, we could even avoid requiring Windows though.  I would
prefer to see something more web-based -- e.g. a PHP script which
contributors could upload to their linux servers which takes care of all
this stuff, using a web-app as the GUI.  (virtually every potential
mass-check contributor can install a PHP script competently, I'd say.)


defining too many implementation details this early may be overkill
anyway.  see http://wiki.apache.org/general/SummerOfCode2006 .

--j.

Reply via email to