[Top] [All Lists]

Re: Last chance to review draft-ietf-sieve-3028bis-10.txt before sending it to IESG

2007-02-06 12:20:10

Alexey Melnikov writes:
> Arnt Gulbrandsen wrote:
>> Alexey Melnikov writes:
>>> I actually do have one issue with a recent change in -10 which I
>>> consider to be serious, because it broke my Sieve implementation
>>> (as well as CMU's)
>> Why did this break your implementation? It did not cause any
>> previously valid scripts to become invalid. All it did was clarify
>> an existing rule.
> If somebody puts "100k" in a Sieve script, Sieve lexer is going to
> reject such script.

Yes. But that's not lasting damage - the user has to change kmg to KMG
and upload the script again.

Compare that with the alternative: Scripts which have passed
managesieve's putscript checking become syntactically invalid when the
server is upgraded.

>>> I actually have a problem with *numbers* being case-insensitive.
>>> Cyrus Sieve doesn't accept "k", "m" and "g" prefixes in numbers.
>>> So I would like to poll the WG regarding this issue. Do people want
>>> to make to make number case-insensitive, or should this change be
>>> undone?
>> Making numbers case-sensitive could invalidate extant valid scripts.
>> That needs strong justification IMO.
> Let's assess the damage first. If it turns out that the majority of
> implementations implement case-insensitive suffixes, then I will have
> to fix my implementation ;-).

I wrote my code case-insensitively, but it's unreleased as yet.

FWIW, our implementation also handles either case. The only "problem" we've ever
had with this feature is users who enter values like 100000G and then get upset
when it says "integer too large" or something similar. (Yes, we frequently get
absurd complaints like this.)

I believe our script generator puts these out in upper case and I know I've
seen hand-constructed scripts have them in lower case. So scripts are probably
gonna break if we make this case-sensitive.