[Chicago] GitHub & SpamBayes

[Skip Montanaro]

On Mon, Feb 4, 2013 at 8:12 PM, Adam "Cezar" Jenkins
<emperorcezar at gmail.com> wrote:
> I think you're mistaken on how most people find projects and contribute to
> them. In my opinion most people find a project because of a need.

That I most certainly understand.  Those of us who are/were SpamBayes
developers worked on it because it scratched an itch we had.  It's
still in use on mail.python.org (whether you realize it or not - it
provides a bit of spam filtering for most mailing lists like this one,
and serves as the only spam filtering for Usenet news messages
gatewayed to python-list at python.org).  We all used to use it.
Eventually though, our mail service providers wound up in one of three
camps:

* Some developed good spam filter technology themselves.  Gmail falls
into this category.  I presume other major big name mail providers do
as well (Yahoo, Hotmail, etc), though I have no direct experience with
them.

* Others were pressured by the above 800-pound gorillas to not simply
forward all the spam along.  Pobox.com, who provides mail forwarding
for me, falls into this category.  It was easy to provide leverage.
All the big guys had to do was threaten to start refusing mail from
those forwarding services if they didn't do what they could at the
front-end (while the SMTP connection was still open - blackhole lists,
greylisting, etc).

* They were driven out of business because they didn't do what groups
like pobox.com did.

I wound up with effectively no spam reaching my mailbox, so I had
nothing for SpamBayes to munch on.  I eventually decided that the
combination of Gmail + pobox.com was "good enough".  I suspect most
other open source developers fall into the same camp.

SpamBayes is now left with a few users who need help and are decidedly
not developers.  I doubt that most of the open source developers who
could fix the problems SpamBayes has on Windows would need its
features either.

> If it
> doesn't work and they are developers, they will probably want to make it
> work. Github makes it much much easier for them to push a button and push
> those changes back to you.

I find it unlikely that a developer for whom SpamBayes scratches no
itch is going to fix bugs.

> Without something like Github, the barrier to contribute back is higher.

But isn't that Git speaking, not necessarily GitHub?  Could I just
convert the Subversion repo at SF to Git and get the same benefit?

> Your original post talked about the users of your project, so you already
> have a community. You talk about how they need to use the project, so you
> already are fullfilling a need. What you don't have is a simple way to
> contribute back. You had a user suggest Github as a way to do that.

Yes, but the people with the problems are not developers.  The
SpamBayes developers no longer need or use the tool.  I'm left with a
community of users who can't help themselves (well, except perhaps for
this one fellow who didn't say, "move to GitHub and I'll fix your
problems", but "move to GitHub and someone else will fix your
problems"), and a community of (former) developers who no longer use
the tool and have moved on to scratch other itches.

It's unclear to me that the investment of my time in moving from
Subversion to Git (be it at GitHub, SF, or somewhere else) is going to
pay itself back.  It would be different if I was still actively
working on SpamBayes myself.  Then I would have no problem moving to
Git (or Mercurial, or Bazaar).  With no current developers, it would
seem to make little sense to move the code repository on the
off-chance that someone will find it and start contributing.

Skip