/ Mark Nottingham <mnot@xxxxxxxx> was heard to say:
| On Jul 12, 2004, at 7:03 PM, Tim Bray wrote:
|
|>
|> On Jul 12, 2004, at 6:15 PM, Dare Obasanjo wrote:
|>
|>> Those rules seem fine with me except they don't
|>> account for backwards incompatible changes. This is a
|>> key part of any versioning story.
|>
|> For incompatible changes, I really only see two options:
|>
|> 1. Provide a way to signal MustUnderstand so that back-rev software
|> can fail gracefully.
|> 2. Use a different namespace.
|
| Can you spell out option #1 (or give a reference into this sea of bits
| we call the mailing list)? On the face of it, I don't see how a mU bit
| is adequate on its own (it makes a lot of sense in conjunction with a
| different namespace).
In a nutshell, if an application sees:
<atom:undefinedName atom:mustUnderstand="true">xxx</atom:undefinedName>
then it must halt and catch fire. If it sees:
<atom:undefinedName atom:mustUnderstand="false">xxx</atom:undefinedName>
it just ignores the element. I feel pretty strongly that the default
value of atom:mustUnderstand should be "false", but it's an open
question.
Be seeing you,
norm
--
Norman Walsh <ndw@xxxxxxxxxx> | Endurance is frequently a form of
http://nwalsh.com/ | indecision.--Elizabeth Bibesco
Attachment:
pgp00120.pgp
Description: PGP signature