detect and allow the default Accept header sent by HTTPURLConnection

[romain-grecourt]

Fixes #211

[spericas]

Wouldn't it be simpler for our media type parser to interpret "" as "/*" instead of matching an exact (and bogus) header?

[romain-grecourt]

@spericas The initial idea is to hardcode a workaround specific to HTTPURLConnection and avoid allowing such bogus headers (*; q=.2) in general.

I'd be fine either way, what do others think ? @tomas-langer @ljnelson @barchetta

[barchetta]

I view this fix as a work-around for the specific JDK bug. So I'm OK with the narrow fix as proposed.

If we later learn of other clients that violate the spec and use a single *, then it would make sense to follow Postel's Law and update the fix to be more liberal in what we accept.

[spericas]

@spericas The initial idea is to hardcode a workaround specific to HTTPURLConnection and avoid allowing such bogus headers (*; q=.2) in general.

Just to be clear, it wouldn't be *; q=.2 but simply* regardless of whether there are params like q or not. I just find the addition of the conditional,

acceptValues.size() == 1 && HUC_ACCEPT_DEFAULT.equals(acceptValues.get(0))

a less elegant and more fragile solution. OTOH, it may be sufficient.

[romain-grecourt]

@spericas Yup, sorry I was misleading with "*; q=;.2", I did that to highlight the part of the header sent by HUC that is bogus. I agree that the current solution is not elegant and is fragile, however the main goal is to workaround that specific JDK bug.