BERP update

[Lucky, the tourist]

Before BERT-RPC was released I had created a small network project
that encodes messages with a similar format to BERP. The header was 5
bytes: 4 bytes for length and 1 byte for flags which included gzip and
error. For BERT, I propose the following header format for the next
version (1.1? 2.0?).

1-byte BERT version (char.char)
4-byte length
1-byte flags, {gzip, ...}

There is an implication with regards to backward and future
compatibilities. A BERT 1.0 client will incorrectly parse a newer
version of the format, unless we moved the version *after* the length.
However, if we want to have a larger possible length for messages, we
need to know the version number before we parse the length. What do
you guys think?

For my BERT usage, I need to pass around a ton of data and having the
option to gzip the data would help tremendously, especially if I don't
have to modify the server to do so. I'm sure others could find more
uses for flags--my brain capacity hasn't returned yet (just got back
from PyCon).

Any feedback would be greatly appreciated.

[Benjamin Black]

On Mon, Feb 22, 2010 at 1:00 PM, Lucky, the tourist
<luckyth...@gmail.com> wrote:
...

> For my BERT usage, I need to pass around a ton of data and having the
> option to gzip the data would help tremendously, especially if I don't
> have to modify the server to do so. I'm sure others could find more
> uses for flags--my brain capacity hasn't returned yet (just got back
> from PyCon).
>
> Any feedback would be greatly appreciated.
>

Use HTTP.

b