Subject: Re: Free software as a replacement for Microsoft
From: Tom Lord <lord@regexps.com>
Date: Sun, 16 Sep 2001 00:39:12 -0700 (PDT)



Larry,

Can you withstand a caustic reply?  I thought you could...

Software architecture *does* matter.  Deeply.  MS's architectural
commitments are their weakness, because they constrain what
capabilities may be practically offered to users.  Sadly, for us in
the Free Software world, MS executives choose to heavily invest in R&D
aimed at relaxing those architectural constraints.  To a first
approximation, today's Free Software corporations don't invest
similarly -- why, they're downright hostile towards it.

I agree that the less-technical users' perspectives are critically
important; I agree that a lot of open source projects on the current
landscape neglect that perspective.  However, you seem to imply that
any MS competitor must match MS feature for feature, which I think is
plainly foolish.  Rather, any MS competitor must match (and beat) MS
capability for capability, regardless of what feature set is used to
provide those capabilities.  

I'm pretty sure, to borrow your example, that you can manage mailing
lists on unix without reimplementing Excel.  Heck, you can probably
even print address labels.  The real question is how to put that power
in the hands of less technical users.

In racking up capabilities, MS has a software architecture that works
well if you have an army of compliant programmers to direct.  You can
decree an architectural tweak at the seat of power, and the army
applies that tweak to many different pieces of code.  In this way, all
the different components are perpetually, increasingly, interdependent
and "intertwingled".  There's no chance at all of a platform competitor
coming in to replace just one core component (and thereby gain
leverage).  That's bad software engineering -- but it sure helps
maintain a state of lock-in for both users and third-party developers.
Of course you switched to IE.

Other architectures are possible that are better suited for unix and
open source development processes.  It's disappointing to see so much
open source effort go in the direction of parroting MS's approach. At
the risk of self promotion, I recommend reading about software basis
sets and their spans in http://www.regexps.com/nature/nature.html.

In my experience, a serious problem in our industry is that too many
executives don't give a demonstrable whit about software architecture.
They're curling.  The R&D labs of yore set the stones in motion and
today's clever Free Software engineer-employees are pretty much all
working as sweepers.  MS, playing a different game entirely, is
skating right past them and picking up speed.  See Bill.  See Bill
skate.  See Bill skate very, very fast.  Skate, Bill, skate!

Regards,
Tom Lord

p.s.: I've seen plenty of situations where even the reduced feature
set of today's Linux desktop software is more than sufficient.  I'd
never recommend it though -- too many bugs, too little attention to
detail, and no process in place to fix that.  How unfortunate that at
least the publicized effort is focused on expanding the architecture
and feature set.  I guess today's FSB execs aren't too sharp on
tactics, either.