[MLton] mlton.org server

[Matthew Fluet]

Steve,

On Tue, Mar 1, 2011 at 6:13 PM, Steve Sims <sims at reactive-systems.com> wrote:
> As you both know, MLton is very important to our company, so if there is
> anything we can do to assist with some of these logistical issues we would
> be happy to.  This might be in the form of work on tasks we might be able to
> perform such as work on the website or financial contributions if you decide
> to go with paid services rather than something free such as SourceForge.

Thanks for the vote of support!

> I guess these changes would have a couple of primary goals in my mind:
> 1)  Remove administration burdens from MLton developers so the time they
> have can be focused on the fun stuff of actually working on the compiler.
> 2)  Better promote MLton so that more people learn about it, try it, use it,
> and want to contribute to it.

I think that the goal is more heavily weighted toward (1) than (2),
though anything that contributes towards (2) would be good.

> I do not have first hand experience as a user or administrator of
> SourceForge, so if you have used it and liked it then most likely it would
> be good for goal 1 above.  I'm not sure about goal 2 however.  It has always
> been my impression that SourceForge is so focused on developers that it does
> not present well to the outside world.  My perception is that a project has
> more credibility if it has its own website (e.g. mlton.org) verses a project
> page on SourceForge (http://sourceforge.net/projects/mlton/).  Perhaps I
> just do not understand how it works though.  Is there some way to use
> SourceForge to stealthily host mlton.org as opposed to appearing as a
> project within SourceForge?

I believe so.  Indeed, as you noted, Stephen moved his homepage to
Google sites, yet it advertises itself on the URL bar as "sweeks.com".
 But, your point is well taken.

A little history: Prior to 2002, MLton's public presence was hosted
under sourcelight.com (which was different from the current
sourcelight.com) and before that under research.nj.nec.com.
web/cvs/mail moved to SourceForge in April 2002
(http://mlton.org/pipermail/mlton/2002-April/021730.html); by
September 2002, web had been moved to another service
(http://mlton.org/pipermail/mlton/2002-September/022332.html); one
reason was not being able to get web statistics.  It was in October
2003 that web/cvs/mail were moved to a new machine
(http://mlton.org/pipermail/mlton/2003-October/024493.html); some
disruptions in cvs and mail services were the motivating factors.

So, it is true that we weren't entirely happy with SourceForge.  And
it is hard to say whether the situation would be significantly
improved now.  There haven't been (m)any votes of confidence in
SourceForge by those who have responded to the thread.

> If you wanted to explore alternatives, the first question would be whether
> you go with paid services or free services.
> For free services (web option), Google sites is pretty good.  I see that
> Stephen is now hosting his personal site there.
> There are a lot of inexpensive web hosting options also.  If you prefer the
> content-management-as-a-service route, squarespace.com offers something
> somewhat like Google sites -- you do all the updating of the website from a
> web browser.  Alternatively, you can roll-your own.  Our company uses
> linode.com who offers a linux virtual machine starting at $20.  Under this
> route you manage the web server yourself.   We of course use SML for our
> content management :)  We use ML Server pages for our external site
> (www.reactive-systems.com) and SMLServer for our internal intranet.  Both
> work great.  Once it is set up, its pretty easy to maintain, and of course
> it would be good PR to be using SML to implement the site.   I can give more
> details if you decide to consider this route.  If you do, we could help get
> it set up and with the porting of the current site (which as Matthew points
> out is not going to be trivial no matter which new framework you go with).
> As for svn and email discussion lists, as I said I don't have experience
> with SourceForge, so that might work fine.  There are commercial
> alternatives.  We host our own svn repository, but we had done a bit of
> investigation of companies that offer svn hosting.  There are a bunch of
> them, some very inexpensive.  For Mailman hosting there seem to be some, but
> not as many providers.  From some quick Google searching, I could not
> determine if the hosting services allow you to have the archive on your own
> site.  In my mind this would be the biggest advantage of such a service over
> SourceForge.  That is the archive would just appear on mlton.org (as it does
> now) instead of at SourceForge.

All good points.  I believe that we are essentially in the linux
virtual machine solution right now, administering our own webserver
(including wiki), mailing list, and svn.  I don't believe that it was
difficult to setup, but its the maintenance 8 years later that is
becoming problematic.  Perhaps its simply that there wasn't enough
maintenance in the intervening time, but I think that is part of
Stephen's "5-year timeframe" requirement --- it is hard to predict
when someone will need to shift focus.

For example, mail right now is the biggest "administration".  Perhaps
it made sense at the time, but there is a white list between the
{mlton,mlton-user}@mlton.org addresses and the mailman queues.  That's
why you have occasionally seen "Resent by:" e-mails from me, as I
discover e-mail addresses that weren't on the white list.  Meanwhile,
*everything* filtered out by the white list is sent to Stephen and
myself (for me, to this GMail address), but this has caused GMail's
spam filter to learn that nearly everything sent to me by way of
*@mlton.org is spam; as a consequence, I've missed some legitimate
e-mails and requests.  And as much overhead as that is, it seems to
pale in comparison to deciphering the mlton.org mail setup and
changing it.  In any case, my vote going forward is to switch the
mailing lists to subscriber-only posting.  I think the anyone allowed
to post decision was motivated by the fact that we ask bug reports to
be sent to mlton at mlton.org, so it made sense to make sure that such
reports were received and not bounced back.  That would suggest that
we also go to an issue tracker of some form.

Also, I do think the mail archives are important.  Very often, I point
back at old mail discussions to explain/justify design decisions.
Though, if we weren't happy with the web-archive of a mailing list
service, as long as we had access to the raw archive, then we could
recreate the "http://mlton.org/pipermail/" structure.  Indeed, it
might be a good idea even if we were happy with the web-archive of a
mailing list service.

In my mind, another means of reducing administrative overhead is by
consolidating on one hosting service, rather than distributing around
over multiple services.

> This topic is related to something I have been thinking about for a while,
> namely the promotion of Standard ML in general through standardml.org.
>  Stephen, I see you are listed as project manager there.  Are you at all
> involved with that any more?   Since it has not been updated recently, I
> worry that newcomers might mistakenly get the impression that not much is
> going on with the language any more.  I think that it is really a selling
> point of MLton that it implements a formally defined language that has a
> number of high-quality compilers that (more or less) conform to the same
> standard.  So it seems to me that promotion of Standard ML in general is
> also good for MLton.
>
> Would it ever make sense to form something analogous to the Apache
> Foundation to promote Standard ML?  Perhaps the "Standard ML Foundation"?
> The idea would be for it to provide the types of infrastructure support that
> this thread is discussing for the different SML-related projects (different
> compilers, standardml.org, etc.).  For each of the services brought up (web,
> email discussion list, svn), one account would probably be adequate to
> support a number of different SML projects.   A foundation would provide a
> vehicle for companies such as ours to cleanly donate towards this type of
> infrastructure.   The company would give the money to the foundation and
> then the foundation would purchase the services to support the various
> activities.  I believe some of the service providers also offer discounts to
> non-profits.

Certainly an interesting idea.  As much as I appreciate the importance
of the social/community side of things, I must say that the whole
engineering of such is a mystery to me.  Not to mention the whole
legal/logistical side of such a foundation.

> Whatever you guys decide we would like to support you how we can.  We
> appreciate the work of all the MLton developers.

Thanks.