On Sunday 15 May 2005 15:55, Peter Stuge wrote:
> This is a very interesting philosophical discussion that is becoming
> increasingly common with many open source developers sharing your way
> of reason. From a marketing perspective it's suicide, however, which
> is also important to consider.

Not really.  Look at the amount of 0.x version software on any given linux 
distribution.  I use 0.x software on my desktop machine, I use 0.x software 
on production servers for years on end without an upgrade, and so on...

For one, I don't think anybody's trying to market Binc.  Yeah, it's always 
nice to have a few more users, but it's additional headache too.  I don't 
know any open source developers who have god complexes like Bill Gates and 
want everybody on the planet to use their product.  Rather the generally just 
write because it's what they love to do, and make a product according to 
their philosophies on the subject because they want to.  They find a few 
users to test it and try it out, and pretty soon, they suddenly have a 
userbase :).  Sometimes this scares the developer away into hiding for years 
on end :/...

What really matters is the user perspective, you say.  Well, looking at the 
industry over the last decade or so, I would say that your argument would 
have been a lot more valid in the past than it is now.  All the big companies 
have made a joke of version numbers, even more than those who made a joke of 
them before did previously ;-).  Windows went from marketing-oriented normal 
version numbers, to year, to 2 digit codes, back to years, back to 2 digit 
codes, back to year, back to yet another random 2 digit code (this time 
starting with an X because that's what's popular marketing these days), and 
what's this?  The latest release uses a year again, but the next one to come 
out doesn't.  To make matters more confusing, the particular branch that used 
years first switched to letters, and the other branch which used 2-digit 
codes first (NT) switched to using years.  Maybe we can get Andreas to write 
a POP3 server and he can trade the version numbers between them every few 
months ;-).  Microsoft isn't the only company who's done this, Apple jumped 
on the X bandwagon and has not released a new version since, even though some 
of their point releases have been quite different enough to normally require 
a major version change.  Red Hat started jumping version numbers for the sole 
purpose of marketing.  This was mocked by insiders, and was the reason 
Slackware jumped it's version from 4.x to 7.0 without ever releasing 5.x or 
6.x (as stated in the release notes).  Speaking of Red Hat, they were about 
to jump on the X bandwagon, but by then Athlon was already naming their 
processors after operating systems and X was used for so many things that it 
was starting to be recognized for the marketing ploy it was, so the stopped 
after 9 and went back to 1.  Yes, I'm exaggerating slightly, but I think you 
can see my point.

As an example, when I decided to move my production mail servers from BincIMAP 
to Dovecot for the time being, I never even thought about version numbers.  I 
thought "I need an IMAP server, one which works well with 
qmail/vpopmail/maildir, one that supports all of the important stuff my 
current one does, and one that overcomes the problem I'm having with what 
I've got.  So I went to google and browsed around for a while.  I've used 
courier before - I now know better.  I then came across Dovecot, read a bit, 
and gave it a shot.  Did it ever even cross my mind that it didn't have a 1.x 
release out yet?  Nope.  I didn't even realize the version number until 
Andreas said not to use 1.0-test and I looked at the version number I had 
installed.  Also of much greater importance than the version number is how 
close the philosophy of the author is to my own (which is the main reason 
I've been using bincimap for the last couple years).

It's also become much more of a trend to avoid early releases in a new major 
version release (or minor version if major version 1 has not yet been 
released) when stability is of importance.

To go even farther, I trust high-version packages (i.e. BIND 9.x) much less 
than lower-numbered versions (i.e. djbdns 1.05).  That's just become 
instinctive over time as I've used different software and realized that stuff 
that's been around for too long tends to get more bloated and old bugs and 
bad programming don't die easy.  I don't have any logical reasoning as to 
why, but this change almost always happens after version 4.  Version 4's seem 
to always be really quite nice.

> To come back on topic I'm glad Binc is past 1.0 and perhaps 1.3
> should become 2.0 rather than 1.4? Do you have a release roadmap with
> features planned for Binc, Andreas?

I think it's way too early for 2.0.  I don't know how close Binc was, but my 
opinion is that the 1.0 release should be reserved until the application is 
"completed", meaning that the author feels that it has become the masterpiece 
he set out to create when he started.  Of course, when he looks at this 
masterpiece, he probably sees room for improvements or has ideas on how to 
redo parts of it, but it is nonetheless a complete implementation, known by 
extensive testing to be very stable, with all known bugs fixed, no matter how 
difficult or obscure (feature requests are another story).  Major version 
number changes should be reserved for when the product has undergone *major* 
rewriting, *severely* breaks compatibility, or has just become so much of a 
different product that it could just as easily be released under a new 
name ;-).  Generally, one would try to maintain compatibility and easy 
upgrades from version 1.anything to the latest 1.x release, and save up all 
those big system-breaking changes in a TODO until there is sufficiently 
enough to qualify for the previously listed qualifications (and the author 
also feels that these big system-breaking changes are done for the best, and 
he probably spends lots and lots of time going through the code and making it 
better by breaking it even more ;-) ).  I don't know how far along that path 
Andreas thinks Binc is, but I would say that in the case of bincimap, 2.0 
should wait until completion of support for all the standard features and 
imap extensions, that is if Andreas lives that long and does not commit 
suicide from having to deal with A> our bothersome banter and B> IMAP clients 
such as Outlook 97.

(and yes, I think we're pretty off topic ;-) )

Cheers,
-- 
Casey Allen Shobe | http://casey.shobe.info
[EMAIL PROTECTED] | cell 425-443-4653
SeattleServer.com, Inc. | http://www.seattleserver.com

Reply via email to