Re: [I18ndir] Working Group Last Call: Structured Headers for HTTP

["Patrik Fältström " <[e-mail]>]

On 4 Feb 2020, at 17:06, John C Klensin wrote:

> Does that make sense?  Do we think that sort of requirement is needed and would be helpful?

Yes, something at least. Because from my point of view IF you use Unicode in something that is to be compared as a protocol element (header and/or value) one should specify what the comparison algorithm is. This might be that data is to be normalized or that normalization is to happen at the time when comparison happens.

Just not stating anything -- even if the field itself is ascii only as in this case -- is I think a bit naive. Just at least say something. Which in the case of "headers which are ascii only while values can be UTF-8 encoded values" could be very easy.

I just want "comparison" be mentioned when people talk about unicode. Regardless of how they then specify the comparison algorithm.

Comments on the comparison algorithm specified is a different thing.

Am I wrong?

I just see too many programmers moving from text strings to unicode strings and then they think they are done. I am so hard trying to tell them "just do one more step" and think about comparison (even if the library they use "might do the right thing") for a few seconds.

I have been burned too much for example with the NFD normalization of the Apple file system which forced use of special rsync software which of course was newer than what Apple distributed etc... We all have been down this hole.

   Patrik