use of "word" ABNF production

[julian.reschke@…]

We should be more consistent in using "word" (= token / quoted-string).

Either remove it, or use it always.

[fielding@…]

From [2519]:

(editorial) Remove unnecessary or duplicate ABNF that can be replaced with core rules: word, specials, attribute, value, and quoted-cpair; addresses #541

[julian.reschke@…]

From [2520]:

re-add 'special' ABNF production that was removed in [2519]; it's there to make clear what characters are indeed special wrt to tchar (see #541)

[julian.reschke@…]

From [2521]:

change tracking for [2519] (see #541)

[fielding@…]

I don't understand why special was restored in [2520]. I know why it was in the spec originally, but it no longer serves that purpose. It isn't used as such in the spec, isn't used in practice (parsers), and isn't useful by external specs.

If the only reason is to explain the set to folks who don't read ABNF, then we should do that using a better comment (or better prose). For example,

     tchar          = "!" / "#" / "$" / "%" / "&" / "'" / "*"
                    / "+" / "-" / "." / "^" / "_" / "`" / "|" / "~"
                    / DIGIT / ALPHA
                    ;
                    ; VCHAR, excluding "(),/:;<=>?@[\]{}"

[julian.reschke@…]

The main reason I restored it is that we had zero discussion before you removed it, and we are past IESG evaluation.

I believe it is useful to have it in the ABNF for clarity; but moving it into a comment works for me as well.