W3C home > Mailing lists > Public > public-appformats@w3.org > September 2007

Re: [ ACTION-107] Updating version list identifier algorithm

From: Marcos Caceres <marcosscaceres@gmail.com>
Date: Mon, 10 Sep 2007 17:48:16 +1000
Message-ID: <b21a10670709100048r75211274ub817c8051bc6429e@mail.gmail.com>
To: "Arve Bersvendsen" <arveb@opera.com>
Cc: "public-appformats@w3.org" <public-appformats@w3.org>

Hi Arve,
> Comparing version lists
...
> 2. Compare the list items n[p] and m[p] using a natural sort algoritm [1]

I've been reading over the Unicode Collation Algorithm (UCA) [1] and
I'm getting the feeling that requiring implementations to support UCA
(or something similar) would be overkill for version information:
localized strings comparison in an internationalized context gets
quite complicated as you would need to identify which language the
version information is written in, etc, to do it properly. It also
opens up a big can of worms about internationalization support for
just one minor area (versioning), and not for others
(resources/content adaptation).

Even though I originally pushed for having strings in the version
identifier, after reconsideration I think we should drop back to the
original proposal of just using non-negative integers delimited by a
"." (as Firefox, and Yahoo!'s Widget engine currently does):

eg. 0.1, 1.0, 1.101.03, etc

If people really want string-based versions lists (eg. "Version 1.0 -
(Beta1)"), we can probably add it in version 2 of the spec.

Any further thoughts?


[1] http://www.unicode.org/unicode/reports/tr10/
-- 
Marcos Caceres
http://datadriven.com.au
Received on Monday, 10 September 2007 07:48:24 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:10:22 GMT