Skip rendering of template by the autopage handler on HTTP methods other than GET

[racke]

Tiny optimization for autopage handler - it doesn't make sense to render the template and throw it away.

[veryrusty]

rfc 1945 - HTTP 1.0 section 8.2 and (the obsolete) rfc 2616 - HTTP 1.1 state

The metainformation contained in the HTTP headers in response to a HEAD request should be identical to the information sent in response to a GET request

i.e. A HEAD request should include a Content-Length header.

Whereas rfc 7231 - HTTP 1.1 section 4.3.2 allows Content-Length header to be missing.

After #1139 was merged, the remaining logic in the AutoPage handler this Pr is optimising does need some cleaning up. However, I think we should always render the content and let our middleware stack remove the body for a HEAD request, so the Content-Length header is included in the response.

[veryrusty]

and .. @racke++ for picking up that the logic needs cleaning up there!

[xsawyerx]

I think we should benchmark not using the Head middleware. If we're already at the spot where we have the body and headers and we can just remove the body, that will save at least two subroutine calls per request for a single if() condition. It's much better.

(Some of the optimizations I did in D2 to make it faster than D1 was removing unnecessary middlewares - it made a big difference.)

[xsawyerx]

So, so summarize:

• I agree with @veryrusty on the spec. HEAD should be GET minus the content, meaning we do need to run the GET, just not return the content.
• I will open a different ticket to remove the HEAD middleware. Middlewares are more expensive than we assume.

Should we close this issue?

[xsawyerx]

Done - #1170.

[cromedome]

Doing some housekeeping... looking at this PR, the ensuing discussion, and #1170, is it ok to close this? I am inclined to. Thanks!

[SysPete]

@cromedome agree: time to close this one.

[cromedome]

Thanks!