Front page

Sereal encoding scheme

0a7d5c2a6bcf49a3ad01dde1996a35b9
ALLOW INSINCERE ESCAPE

From: Gabriel Moreau <Gabriel.Moreau@legi.grenoble-inp.fr>
Date: Tue, 1 Nov 2016 15:56:00 +0100

   Hi,
   
   On the web page :
   
     http://obnam.org/format-green-albatross/
   
   I read :
   
   "
   Objects are stored in a suitable serialisation encoding.
     Instead, a simple, custom binary encoding will be used...
   "
   
   I remember some work about the Sereal protocol done on Perl.
   
     http://search.cpan.org/~yves/Sereal-3.015/lib/Sereal.pm
   
   Sereal seem fast, well documented with somme tools
   
     https://github.com/Sereal/Sereal/blob/master/sereal_spec.pod
   
   Maybe reuse the Sereal encoding scheme could be a good idea ?
   
   gabriel
From: Gabriel Moreau <Gabriel.Moreau@legi.grenoble-inp.fr>
Date: Tue, 1 Nov 2016 16:33:34 +0100

   > Having code in Perl is of little use for Obnam, which uses Python. If
   > there's an implementation in Python, it's not in Debian. I don't want
   > to add dependencies outside of Debian to Obnam.
   >
   > Also, http://blog.liw.fi/posts/dependency-cost/ is relevant.
   
   I agree with you. Sereal was started in Perl but could be use by another 
   language today
   
     https://github.com/Sereal/Sereal
   
   But you right, I saw nothing in Debian about it outside Perl.
   
   > Given the Obnam implementation has been written since May, 2015, and
   > doesn't seem to have any issues I care about (and that means it's not a
   > speed bottleneck either), I'd really rather not start looking at other
   > ways of doing the same thing.
   
   No problem. It was just an idea after a few days break in France ;-)
   
   gabriel
From: Lars Wirzenius <liw@liw.fi>
Date: Tue, 01 Nov 2016 17:07:09 +0200

   On Tue, 2016-11-01 at 15:56 +0100, Gabriel Moreau wrote:
   > I remember some work about the Sereal protocol done on Perl.
   
   Having code in Perl is of little use for Obnam, which uses Python. If
   there's an implementation in Python, it's not in Debian. I don't want
   to add dependencies outside of Debian to Obnam.
   
   Also, http://blog.liw.fi/posts/dependency-cost/ is relevant.
   
   > Maybe reuse the Sereal encoding scheme could be a good idea ?
   
   Given the Obnam implementation has been written since May, 2015, and
   doesn't seem to have any issues I care about (and that means it's not a
   speed bottleneck either), I'd really rather not start looking at other
   ways of doing the same thing.