[volunteers] [svlug] Linux to Politics ratio

Rick Moen rick at linuxmafia.com
Sun Dec 2 00:21:01 PST 2007


Quoting Daniel Gimpelevich (daniel at gimpelevich.san-francisco.ca.us):
> On Sat, 01 Dec 2007 18:58:54 -0800, Lisa wrote:

> > I disagree.  My opinion is that, if the mail and the webserver
> > are on two different machines we are *better* off because if the
> > most unreliable machine we have dies, we still have the
> > webserver running (where we can explain why the mailing list is
> > down) and, if linnode crashes (doubtful that that would happen
> > for long), we'd have the mailing list to explain why the website
> > is down.  Whereas, as it is now, if the most unreliable machine
> > we have dies, we lose BOTH.
> 
> No, because while the new website would not be a souped-up reinstallation
> of the old website setup, the mailing list stuff would be pretty much that
> at first and is in more immediate need of migration than the website. Once
> the Linode host takes over in exactly the capacity in which the current
> host is serving, the current host may be repurposed for a time as the new
> home of the website, and the multiple-basket scenario you're preferring
> could still happen. Would this still be consistent with the agreement you
> negotiated?

The moves of (1) all mail including Mailman and (2) non-Mailman webstuff 
off the old svlug.svlug.org host are entirely _orthogonal_ issues.  The
only reason we were tackling non-Mailman webstuff[1] first is that it was
more difficult -- especially with the wikification requirement that the 
(annoying, dumb) plan inherited from J. Paul Reed necessitated!  Moving
the mail involved nothing but familiar software and seemed easy.  Let me
go over this yet again:


The plan agreed to at Black Angus was, therefore, Phase One, to
replicate the Web site (less Mailman Web pages) at Linode -- except,
substituting on the fly the MoinMoin conversion.  I _hoped_ we would
de-complicate that task by scrapping J. Paul Reed's pointless[2]
insistance on wikification (now that he'd left office), but somehow my
reminders that the justifications for the project _made no sense_ fell
on deaf ears, so to my dissatisfaction and dismay we proceeded with
Reed's wikification plan.

Note that Lisa Corsetti took the initiative to _also_ simply copy and
reimplement the existing Web site to Linode -- which I applaud but was
not the (rather bad) plan agreed to at Black Angus.  That instance of
the Web site is available today.  It's done.  It works.[3]  At the time
that Lisa did it, I did not criticise her for it -- it was in fact what
I thought was the sane alternative -- but I did remind her that it
wasn't what we agreed to at Black Angus.

(Attentive readers may spot a pattern in my behaviour:  I have attempted
to implement the rules, e.g., on the mailing list, and the plans that
_have been agreed to_, even if I personally hold different views.)


Phase Two:  Use the VA Linux 2230 box at Via.net to create a Mailman and
SMTP host, copy over the Mailman and Exim mbox files and the Mailman
list definitions and subscriber lists, regenerate the archives
(preferably incorporating for the first time the pre-Mailman majordomo
archive that has been preserved but never webified), and cut over the
DNS for "lists.svlug.org" and "svlug.org" to it, shorting TTL in advance
to reduce caching problems.

For Phase Two to be possible, we'll need to regain access to the VA
Linux box, which reportedly got diverted for Paul's secret project at
Heather's house, resulting in our being startled to find that it no
longer responded to ping at its Via.net IP address, and in the need to
track down where it had gone without informing anyone.

> > That would be nice, but I think 1/2 is better than nothing. 
> > Wouldn't it make sense to get some other opinions on this?
> 
> And two halves are better than one.

Daniel, you are new to this conversation, and completely missed the
precding analysis -- the analysis Paul discarded when he threw away what
we agreed to at Black Angus.  But you are fundamentally mistaken in your
apparent belief that there is any reason to tie one project to the
other.

> AIUI, its flakiness is almost exclusively a function of the way it's being
> used and the way it was set up. I was one of the people who chanted the
> wiki mantra, but always in the context of hearing "there's nothing the
> Linode host can do beyond DNS."

I have no idea who was saying that, but whoever that was, was ill-informed.

In January, Lisa got Linode to pump up the allocated resources from 80MB
RAM to 256 MB, and disk from 3072 MB to (IIRC) 8 GB (or maybe 5 GB,
would have to check).  _That's_ a more reasonable machine, though,
actually, my extremely careful work with NSD (for nameservice) and
lighthttp/FastCGI/MoinMoin suggested that we'd be able to operate
reasonably even within 80MB / 3072MB, if we kept Mailman and SMTP
somewhere else.

Micah Dowty, at the point where he negotiated the original virthost,
confidently asserted that _every_ SVLUG service and all of our back
content could fit and operate in the original hardware resources.  As I
attempted to squeeze everything in, I gave Dowty and Reed feedback that
explained why I felt it wasn't going to work.  Ironically, they seemed
to feel I couldn't possibly know what I was talking about, because I'm a
mere professional sysadmin, and they were _coders_!  {pause for irony}

So, they ordered me to make Web + MoinMoin + SMTP + antispam +
nameservice all fit within 80 MB / 3072 MB.  I tried.  It didn't work.

OTOH, the newer Linode resources are actually feasible for all SVLUG services,
if barely.  (I still recommend migrating Mailman + SMTP to a different
box.)

>>> We want to replace the existing static,
>>> log-in-to-edit-files-from-the-command-line website solution
>>> with a wiki for obvious reasons.

This is the J. Paul Reed assertion, and it's _still_ nonsense.  Argh!

> Of course moving the website alone is not worth not doing....

It _is_ worth doing as Phase One.  See above.

Daniel, I hope I'm not coming across as impatient, but this really is
something we've already gone through an extremely large number of times.
It's all over the earlier archives of this mailing list, for example.

> I don't know exactly who rejected that, but remember: Ideas never die.

Ideas (and plans, and agreements) that get ignored might as well be dead.



[1] The existing site was for many years flat HTML/tables with SSIs,
which Lisa recently converted to PHP4/tables.  There are also three CGIs
coded in Perl:  WebFetch, stv, and sponsor_imgs.pl, plus a (dispensible) 
Java applet on the Link Farm page.

[2] Converting to MoinMoin would not be entirely without benefits:  
We would automatically get CSS templating plus truly modern HTML.
Also, people who actually like wiki-style editing environments could
have them; on the minus side, those of us who prefer real text editors
get somewhat screwed.

To get this extremely slender set of benefits, Heather and I had to work
ourselves to the bone recoding all the existing pages in MoinMoin
markup, and re-implementing some features of the existing site remain
unsolved problems.  Thus my continued reminders that the project exists
for fundamentally mistaken reasons.   Reminder:  MoinMoin prototype,
with all of my and Heather's conversion work, is at
http://gemini.starshine.org/SVLUG/ .

[3] At the risk of sounding picky, I believe Lisa ignored my careful use
of minimal software and implemented the site using Apache2 -- with which
she was more familiar, as am I -- rather than lighthttpd.  Given that
the site is RAM-constrained, I think this deviation from standards
should be corrected.  Also, it should be noted for completeness that the
three Perl CGI things and one Java thing are (I believe) not yet
implemented there.





More information about the volunteers mailing list