More and more of the software I’ve been using lately has been open source. With this shift comes a different approach to releasing updates and to version numbering. Following the “release early, release often” mantra canonized by Eric Raymond in The Cathedral and the Bazaar, many major open source projects like Mozilla release “nightly builds” throughout the development process.
The practice isn’t unique to open source. While they don’t release them, Microsoft does nightly builds of developing versions of Windows. The developers of the closed-source RSS reader I use on Windows, NewzCrawler, have be releasing frequent “beta builds” publicly.
I remember an old article in Windows Magazine (I think it was by Fred Langa) that worried that delivery of software over the internet rather than on physical media would lead to “dribble-ware” - software that was let out the door before it was ready, only to be patched after the fact (sound familiar?).
With all of these changes, I’m beginning to wonder if major version numbers are loosing their significance. Three applications that I am relying heavily on these days are all significantly pre-1.0 (Mozilla Firebird – 0.6, Mozilla Thunderbird - 0.2, and Gaim – 0.7). Having been released frequently for months (years) already, I doubt that these applications are any less complete or stable than most “1.0” software.
Comments
not-the-cmax - October 5, 2003 10:38 pm
open source is the way to be - I have had similar experiences with the release numbers being 'almost' irrelevant
Chris - October 5, 2003 11:17 pm
Most of the software I use on a daily basis is either Open Source, or at least freeware. I don't really pay much attention to version numbers anymore, since quite often a 0.2 version of one program is a lot better than some 6.0 version of another :)
NewzCrawler looks nice, compared to some of the RSS/etc readers I've seen, but like the rest, it uses the IE HTML engine. I want one (that doesn't suck) that uses Gecko...
nathan - October 6, 2003 1:01 am
Release numbers in open source are exactly what they used to mean in all software development before marketing monkeys started madly incrementing to beat the competition. And they are still quite relevant as a way of differentiating subsequent releases of the same software. This is their real purpose rather than sounding impressive, creating a buzz about the latest and greatest, or trying to out-number the competion. I'd argue that these other purposes were never really relevant.
Some proprietry software seems to have seperated the marketing numbers from the real release numbers. This probably started around the time the marketing monkeys decided to increment from 3.1 to 95.
Simon Willison - October 6, 2003 7:28 am
1.0 still has a very important meaning in Open Source, it's just different from the meaning expected from proprietary software. In Open Source, 1.0 often means that the interfaces to the main code modules in the project are likely to be "frozen", meaning it's safe to develop your own software on top of the open source project. Prior to 1.0 there is no garauntee that software you develop will still work when the next release of the project is made. Since proprietary software is rarely deesigned to be developed upon in this way this meaning of 1.0 is far less improtant.
Shawn - October 6, 2003 2:49 pm
I would say 1.0 means my mom can use it. It is stable and been through lots of testing, in the case of a project like Mozilla at least. The last thing I need is an ear full from mom a 1000 miles away, over the cell phone, that some build of Firebird I sent her doesn't work with her favorite Soap Opera web-board. As it is I waited till Mozilla 1.2 to get her using that over Mac IE.
As for us computer weenies "1.0" does not matter much.. most of us have been burned by a nightly build of browser X that was rather wonky and I do not think I lost any sleep over it.
Will - October 7, 2003 1:42 pm
What's the attraction to nightly builds over the standard releases? I've been too chicken to play in the nightly build sandbox, because I want my functionality to come with some stability. Am I missing out on something worthwhile?
Tony Bussieres - October 7, 2003 2:23 pm
You might not need of a nightly build, it depends, sometimes nightly builds are unstable,because the implementation of a new feature is not complete yet, or bugfixes are still on the way but not necessairly bulletproof... it also depends how active the testers are.
I'm using myself Nightly & CVS builds, and most of the time it works good. If it doesn't work, you can try to fix it, or advise the communauty around the project and notify your bug. And if you're lucky it will be fixed soon. Behaving that way is a good contribution for the opensource communauty. I like to test the new stuff, like the new Linux kernel 2.6.0-test6 :)
I've found an interresting document about the release practice :
http://en.tldp.org/HOWTO/Software-Release-Practice-HOWTO/
Kevin - October 8, 2003 4:35 am
I'm currently dependant on Firebird and GAIM as well but only update when there is a significant reason to (the y! fix in GAIM for instance). I haven't tried Thunderbird yet, is it a decent email client? Currently I'm using Outlook but may give Thunderbird a try I just hate moving my email around to much.
nathan - October 8, 2003 2:23 pm
Gaim has really been refined nicely in recent updates. I've been updating it whenever possible. BTW, the correct names for the project are Gaim or gaim; all uppcase should not be used to avoid legal problems with AOL.
If you use some of the non-mail features in Outlook, I'd recommend looking at Evolution.