[Top] [All Lists]

Re: What's left before a new RFC?

2002-04-18 07:46:31

Marc Mutz wrote:

Hash: SHA1

On Thursday 18 April 2002 10:17, Marcel Waldvogel wrote:

For the Version and Comment headers, I propose to state that they are
UTF-8, but for interoperability, implementations SHOULD restrict
themselves to generate ASCII characters.


I don't see the how having UTF-8 inside _ASCII_ Armor can be justified. The problem is that the ascii armor is going to be used in non-8but-clean environments. Else, you'd use the binary format, no?

One one side, we can consider cleartext signatures, which already can contain characters with the MSB set, and which are not necessarily armored for 8-bit cleanliness. Even there, the signature block is (among other things) ASCII-armored to not confuse the user (or his terminal) with weird chars.

There are multiple reasons for armoring non-7bit data:
a) The transport does not allow arbitrary 8bit sequences
b) The transport does not allow arbitrary "line" lengths
c) The transport does charset recoding, probably forth and back.

The headers that use UTF-8 are mostly for human consumption (I presume), and most modern non-8bit transports are of the form (b) or (c), which (1) should not harm typical headers nor (2) cause any damage, if they are slightly mangled (IMHO).

BTW: Does the "Version" header contain the product and version used ("User-Agent"), or the version of the standard ("MIME-Version")? bis04 seems to indicate the latter, but usage seems to be the former.