duplicate pasted text in s-maxage

[royfielding]

The commit d4695a3 for #161 added a paragraph to s-maxage that contains the directive must-revalidate. I think this was supposed to say s-maxage and worded for shared caches?

[mnot]

Sorry, been travelling.

This was intentional (although I forgot to update the directive name). s-maxage implies proxy-revalidate, which implies must-revalidate. That seems like a lot of indirection in the definition.

I agree this wasn't a great way to do it, though. I think the options are:

1. Leave it as is (i.e., two levels of indirection).
2. Leave it as is, but use linking more explicitly so that the indirections can be followed more easily (at least in HTML renderings).
3. Inline the appropriate text from must-revalidate into both proxy-revalidate and s-maxage.

I think those are ordered according to my preferences, least to most.

(I strongly suspect many people don't realise s-maxage has these semantics, because of it having a name similar to max-age. Making them inline rather than referenced might help)

[royfielding]

I agree with the conclusion, but am pretty sure that you meant to say s-maxage in that section defining s-maxage. That's all I changed, so it should be fine unless you wanted to add more text to explain why. I don't think it is necessary to explain more because adding s-maxage implies the server wants shared caches to reuse the response.

[mpdude]

I agree with @mnot that the additional semantics of s-maxage are surprising. In particular due to section 4.2.1, s-maxage comes disguised as the "different maxage value for shared caches".

So maxage + public != smaxage, and the difference is the handling of error cases like with stale-if-error?

Could you please try to explain why the additional semantics of s-maxage make sense – that is, why did you not just make it the alternative TTL for shred caches?

Thank you!

[royfielding]

Ah, okay, so the added paragraph had the right directive before but simply doesn’t make any sense given the context.