Best to fix that before we commit to the stable names!

[mem.poly.alloc.class]
[mem.res.mono.buf.ctor]

looking at Annex F for long names, these are far from the only culprits. Seems to be an issue with many stable names from the TS. Compare: fun.wrap.badcall vs. optional.bad_optional_access - we have forgotten how to abbreviate.

Also, underscores seem to make the names look much longer in Annex F - see the Boyer Moore searchers.

Yes, underscores are sometimes replaced with dots instead, e.g. unique.ptr, which looks much better.

[optional.bad.access] or [optional.badaccess] seem better than [optional.bad_optional_access].

I'm also not sure we really need to keep the word "object" in [optional.object.] ... I think it was there to reserve place for [optional.ref.] but we don't have optional<T&> and even if we did we could still use [optional.ref.*] once we add it.

The longest one seems to be [func.searchers.boyer_moore_horspool.creation] in 20.14.13.3.1, and it formats badly in the new xrefs.

@jwakely, @AlisdairM, @zygoloid: We should fix this before we ship the IS!

GB 52 covers this.

Those long names are also the cause for overfull/underfull hbox complaints from LaTeX; see #693.