[e-lang] Incremental CapTP unserialization?

[Mark Miller]

On Feb 12, 2008 7:47 PM, Kevin Reid <kpreid at attglobal.net> wrote:
> Can objects embedded in a message arriving in a vat over CapTP expect
> that:
>
>    - they are all unserialized in the same turn, and
>    - the message is delivered in that turn?
>
>
> Or: is a CapTP implementation allowed to unserialize objects as the
> data describing them arrives, rather than buffering it until the
> complete message is available?

What CapTP in Java for E-on-Java does is:

a) It unserializes the whole message in the same turn, and only after
buffering the complete msg.

b) The unserialized msg is eventually-sent to the target, so the
message is thus delivered in a separate turn.

Since the target may itself be an unresolved promise, I think CapTP
needs to eventually-send the message, so I think #b is necessary.

As for #a, in order to spread a single unserialization across several
E turns, we'd need a major reengineering of Data-E. I don't think this
effort would produce any observable benefit.
-- 
Text by me above is hereby placed in the public domain

    Cheers,
    --MarkM