never_cache should NOT calculate an etag header

[Eloff]

Presumably the purpose of never_cache is to prevent browser caching, so why calculate an md5 hash of the response for etag, which might cause browser caching? Probably etag should be omitted, but if there is a purpose to adding it, make a nice big random etag rather. No point burning CPU cycles AND possibly giving bad behavior too!

Note that the current behavior breaks streaming responses, see #12214.

[Russell Keith-Magee]

You've already commented on ticket #12214, which specifically mentions never_cach, etags and iterator content; there's no need to open another ticket just to posit another opinion on how the problem should be solved.

[Eloff]

Au contraire, if I don't post it as a separate issue people will whine about me posting meta bugs. But as long as someone actually rethinks that crazy code, I promise I won't tell anyone what your meta-bug preferences are.