"Dan Harkless" writes:
>
>Yes, but I still think we should try to fix as many of the known date
>parsing bugs in the new dtimep.lex as we can (cf. my previous mails on the
>subject).
>
I've uploaded some minor tweaks to CVS.
Of the 15659 messages in my inbox, there
were 64 differences in parsing (as measured
by "scan -form scan.time") from 1.0.4.
10 of them were Y2K issues (a description follows).
51 were messages where no timezone was specified. Previously,
nmh defaulted to assuming the message cam from the local timezone,
which I don't believe is a reasonable default. Now they default
to GMT. An interesting side note is that all but one of them
were bulk-mail spam-type emails. Apparently spammers don't like
revealing their location. Perhaps this would an amusing spam filter.
The remaining 3 were not accurately parsed by either 1.0.4
or 1.0.4-dev. 1.0.4 identifed 2 MET's as +0700, although the mail
headers disagree, and MSK (which stands for Moscow, I believe)
as -0700 which it definitely doesn't. 1.0.4-dev treated all three
as GMT, which, while not correct, should be controllable by adding
more mappings. We just need a fairly authoritive source.
Y2K:
There are apparently several failure modes:
1. 00 (should be 2000)
2. 100 (should be 2000)
3. 1960 (should be 2000) 1 message
4. 1969 (should be 1999) 9 messages
5. 1970 (should be 2000)
The first two cases are handled fine. The third
is simply bizarre, and I'm not sure we should need
to deal with it.
support for 4) can be added, but I'm not sure of
it's value. 5) Is the tricky one, since it could
legitimately be a UNIX time. I'm undecided whether
it would be wise to consider pre-1971 dates as Y2K
errors (assuming people are not still sending mail
this year with non-Y2K-complaint MUAs. Since I didn't
recieve any mails of this sort, perhaps we don't need
to worry about it.
Shantonu
*-------------------------------------------*
|Shantonu Sen * (617)225-7269 * [EMAIL PROTECTED]|
|Electrical Engineering and Computer Science|
|Massachusetts Institute of Technology, 2002|
*-------------------------------------------*
