Database connections as subprocesses Lassi Kortela (14 Sep 2019 07:30 UTC)
Re: Database connections as subprocesses John Cowan (15 Sep 2019 01:06 UTC)
Re: Database connections as subprocesses Lassi Kortela (15 Sep 2019 06:28 UTC)
Re: Database connections as subprocesses John Cowan (15 Sep 2019 23:02 UTC)
Re: Database connections as subprocesses Lassi Kortela (16 Sep 2019 08:22 UTC)
Binary S-expressions Lassi Kortela (16 Sep 2019 17:49 UTC)
(missing)
Re: Binary S-expressions Lassi Kortela (17 Sep 2019 09:46 UTC)
Re: Binary S-expressions Alaric Snell-Pym (17 Sep 2019 11:33 UTC)
Re: Binary S-expressions Lassi Kortela (17 Sep 2019 12:05 UTC)
Re: Binary S-expressions Alaric Snell-Pym (17 Sep 2019 12:23 UTC)
Re: Binary S-expressions Lassi Kortela (17 Sep 2019 13:20 UTC)
Re: Binary S-expressions Lassi Kortela (17 Sep 2019 13:48 UTC)
Re: Binary S-expressions Alaric Snell-Pym (17 Sep 2019 15:52 UTC)
Re: Binary S-expressions hga@xxxxxx (17 Sep 2019 16:25 UTC)
Re: Binary S-expressions rain1@xxxxxx (17 Sep 2019 09:28 UTC)
Re: Binary S-expressions Lassi Kortela (17 Sep 2019 10:05 UTC)
Python library for binary S-expressions Lassi Kortela (17 Sep 2019 21:51 UTC)
R7RS library for binary S-expressions Lassi Kortela (17 Sep 2019 23:56 UTC)
Re: Database connections as subprocesses Alaric Snell-Pym (16 Sep 2019 08:40 UTC)
Re: Database connections as subprocesses Lassi Kortela (16 Sep 2019 09:22 UTC)
Re: Database connections as subprocesses Alaric Snell-Pym (16 Sep 2019 11:28 UTC)
Re: Database connections as subprocesses hga@xxxxxx (16 Sep 2019 13:28 UTC)
Re: Database connections as subprocesses Lassi Kortela (16 Sep 2019 13:50 UTC)
Re: Database connections as subprocesses hga@xxxxxx (17 Sep 2019 13:59 UTC)
Re: Database connections as subprocesses John Cowan (16 Sep 2019 22:41 UTC)
Re: Database connections as subprocesses Lassi Kortela (17 Sep 2019 09:57 UTC)
Re: Database connections as subprocesses Lassi Kortela (17 Sep 2019 10:22 UTC)

Re: Binary S-expressions Lassi Kortela 17 Sep 2019 09:46 UTC

John:
> IMO this does not really save anything much over a textual protocol given
> that the servient (server-client) is running (normally) on the same machine
> as the original client.  The pipes are fat.

It's not about performance, it's about simplicity and reliability.

- A format where every object is length-prefixed makes memory allocation
in C much easier: just call malloc() with the right size off the bat and
one read() can fill in the buffer.

- No lookahead is needed so parsers in other languages are also easier.
It generally makes no sense to have lookahead in formats for machines.

- Length-prefixes also obviate the need to escape strings and symbols
escaping, which I would consider mandatory for almost any format meant
for machines.

- Arbitrary binary data is trivial to embed in exactly the same way as
text strings.

I always start from a binary protocol, and ask whether a textual one
saves anything. It's so much easier to avoid all the pitfalls.