(WIP) Return headers as JSON array instead of concatenated string

[JeremyRand]

This PR makes the blockchain.block.headers method return a JSON array of headers instead of a concatenated string, which improves semantic correctness.

[JeremyRand]

Marked as WIP since the protocol docs haven't been updated yet. This change was discussed on IRC last year and there wasn't any objection to the concept; not sure if any concerns might have appeared since that discussion. @SomberNight Is this something we could possibly roll into the protocol bump in #90 to minimize the number of protocol bumps?

[JeremyRand]

(Also happy to squash this PR on request.)

[JeremyRand]

Not sure if the coverage decrease is a dealbreaker, let me know if it is.

[SomberNight]

Ok, we can do this as part of protocol 1.5.
IIRC you want this because namecoin has variable sized headers?

I guess it adds around 162/160-1=1.25% overhead (comma and space?) in the Bitcoin usecase when requesting lots of headers, but that is negligible seeing that we encode raw data as hex strings.

If you document the protocol change I can cherry pick such a commit to #90.

[JeremyRand]

Ok, we can do this as part of protocol 1.5.

Sounds good.

IIRC you want this because namecoin has variable sized headers?

Yes, that's correct.

I guess it adds around 162/160-1=1.25% overhead (comma and space?) in the Bitcoin usecase when requesting lots of headers, but that is negligible seeing that we encode raw data as hex strings.

I think it's close to 3 bytes of asymptotic overhead per 160 bytes of hex (2 quotation marks and a comma; I'd expect no space character since presumably the JSON serializer is optimizing for compactness). But yes, close enough, I think the overhead is negligible.

If you document the protocol change I can cherry pick such a commit to #90.

Yes, I can do that.

[JeremyRand]

@SomberNight Added a commit to change the protocol docs accordingly. I didn't touch the changelog since that would conflict with your branch; let me know if you want me to write a changelog entry.

[SomberNight]

I think it's close to 3 bytes of asymptotic overhead per 160 bytes of hex (2 quotation marks and a comma

Ah I forgot about quotes..

I'd expect no space character since presumably the JSON serializer is optimizing for compactness

Hah. Don't assume, check! I did so now and whitespaces are there.
I've tried to regain the loss caused by this PR by stripping the spaces :P
kyuupichan/aiorpcX#38

[SomberNight]

Please also update the "Example Response" in the doc some lines below.
(no need to keep an example for old protocol; one is enough for the new one)

[JeremyRand]

Please also update the "Example Response" in the doc some lines below.
(no need to keep an example for old protocol; one is enough for the new one)

@SomberNight Done.

[SomberNight]

I have incorporated the code changes into #101.
The doc changes are part of #90.