Re: Re: Dangerous locale patch to 1.1.3

[email protected] said:
> Indeed.  The Right Thing would probably be to look at the character
> sets specified in Content-Type: headers, but then you have to take
> care with multipart messages, and messages with missing or incorrect
> content types are still a problem.  (OTOH, parsing Content-Type should
> also allow saner treatment of non-textual content without requiring an
> extra filter, which counts as a win in my book.) 

I continually wonder whether The Right Thing would be to strip down ifile
to a form that is completely agnostic of text formatting.  A wrapper would
handle all lexing/formatting and simply pass to ifile word/ frequency
pairs.  ifile would not modify the "word" in any way, so it could be
encoded in whatever form the wrapper deems appropriate.  Only requirement
would be that different "word"s have different encodings. This would fit
nicely with the notion of ifile just being a fancy database.  This would
also make it easier to separate ifile into client and server parts.

Jason D. M. Rennie
MIT AI Lab
[email protected]
(617) 253-5339
http://www.ai.mit.edu/~jrennie/