internal/ethapi: remove header.Size in rpc getHeaderByXXX

[jsvisa]

As discussed with @rjl493456442 and also in #27325 (comment). Currently, the PRC of eth_getHeaderByXXX's response will return the header's Size, which depends on the struct Header's size:

go-ethereum/core/types/block.go

Lines 116 to 126 in dc8d91a

	var headerSize = common.StorageSize(reflect.TypeOf(Header{}).Size())
	// Size returns the approximate memory used by all internal contents. It is used
	// to approximate and limit the memory consumption of various caches.
	func (h *Header) Size() common.StorageSize {
	var baseFeeBits int
	if h.BaseFee != nil {
	baseFeeBits = h.BaseFee.BitLen()
	}
	return headerSize + common.StorageSize(len(h.Extra)+(h.Difficulty.BitLen()+h.Number.BitLen()+baseFeeBits)/8)
	}

And this struct's size differs in the different distros(x32 VS x64), so I proposed to remove this field for:

1. getHeaderByXXX is Geth's own RPC, not in standard ones;
2. We have block's size(Guaranteed by consensus), someone may be confused about those two.

[holiman]

I mostly agree with dropping this, but I am unsure why we even have it in the first place. It has "always" been there, and as far as I can tell it has "always" been wrong, using usafe.sizeOf and thus obtaining a platform-dependent value for size.

IMO we should either return e.g. the rlp-encoded size or nothing at all. But IMO needs some more backstory.

[s1na]

getHeaderByXXX is Geth's own RPC, not in standard ones;

Interesting, I thought our "eth" methods match the specs with the exception of eth_call taking extra parameters. Seems you're right, getHeaderBy* are indeed not part of the specs. We should at least document them on our website and possibly move them to another namespace as part of this re-structure

[s1na]

Judging from Header.Size comment:

It is used to approximate and limit the memory consumption of various caches.

It shouldn't be returned from API.

[jsvisa]

but I am unsure why we even have it in the first place.

I know what's going on, this field is changed in 530f78e?diff=unified#diff-c426ecd2f7d247753b9ea8c1cc003f21fa412661c1f967d203d4edf8163da344L939-R980

And the old code is correct, equals to block.Size()

[jsvisa]

We have several ways to do this:

1. keep with the old style, return to the block's size
2. return the header's RLP size
3. just remove this field

My personal preference 3 > 2 > 1

[s1na]

I think 2 doesn't make sense. Since the header has mostly information about the block, not itself. I'm slightly in favor of 1 than 3. We have the field already and block size could possibly be useful.

[holiman]

I know what's going on, this field is changed in 530f78e?diff=unified#diff-c426ecd2f7d247753b9ea8c1cc003f21fa412661c1f967d203d4edf8163da344L939-R980

And the old code is correct, equals to block.Size()

Nope, that commit added the RPCMarshalHeader. The old RPCMarshalBlock still uses the b.Size():

https://github.com/ethereum/go-ethereum/pull/19669/files#diff-c426ecd2f7d247753b9ea8c1cc003f21fa412661c1f967d203d4edf8163da344R994

So RPCMarshalHeader has always been wrong. And in that case, I think we can just remove it.

[jsvisa]

I'm slightly in favor of 1 than 3. We have the field already and block size could possibly be useful.

IMHO, If someone is interested in the size of the block, he has to retrieve it via eth_getBlockByXXX instead of this one. And BTW, if we have to return the block's size, we need to retrieve the block from the backend either, probably not a good deal.

[s1na]

LGTM