[clug] Your Best arguments please

Martin Pool mbp at samba.org
Mon Aug 11 13:53:01 EST 2003


On 11 Aug 2003, Antony Wuth <ajw at pobox.com> wrote:

> Require 3 separate / independent implementations of the protocol / data
> format.

This is slightly silly.  You just banned

 - Ogg Vorbis
 - TeX
 - Postgres and MySQL 
 - rsync (twice)
 - Duplicity

amongst many others.

A definition of open data formats requires discretion and judgement.
As Chris pointed out, merely having something that purports to be
documentation is not enough, in the absence of a guarantee that the
documentation is correct.  Merely being open source is no guarantee
that there is any other way to reach the data other than through a
particular codebase, which may not be maintained in the future.

There is a more assured practical future path for MS Word documents
than there is for Abiword documents.

All of these things are good indicators but none of them are
sufficient:

 - open documentation of the protocol/format

 - managed by a standards body

 - multiple implementations

 - open source implementation(s)

 - transparent formats that can be read at a lower level using other
   tools (e.g. plain text, XML)

 - free of patents or other encumberances

> You would need to have procedures to verify that the
> implementation was actually independent (so MS doesn't just write 3
> different .doc viewers & claim them as independent) On top of this there
> probably should be the requirement that the end client is guaranteed
> future access to at least one implementation of the protocol / file
> format viewing / use code for the lifetime of the data stored / created
> in the system (ie for potential FOI data 30 odd years) 

Guaranteed how?  Who can make 30 year guarantees for computer data at
reasonable cost?

> This would pretty much require at a minimum code escrow (potentially
> expensive)

Not necessarily expensive, but that is a whole other argument.

>  or better still at least one of the implementations to be OSS
> which means the end client could effectively implement their own code
> escrow.

-- 
Martin 



More information about the linux mailing list