gzipped images in SIAP 1.0

[Mark Taylor]

On Wed, 23 May 2007, Norman Gray wrote:

> On 2007 May 23 , at 01.24, Mark Taylor wrote:
>
>> However, ignoring the headers altogether and following Tom's advice to look 
>> at the magic
>> numbers is almost certainly a better idea: more reliable (the headers might 
>> be wrong) and probably easier.
>
> Myself, I would disagree with this, on public-interest grounds.
>
> If a tool responds to breakage by saying `I know what you mean', then the 
> service in question is allowed to continue being broken.  That diminishes the 
> utility of the MIME type system, and forces every other client to guess too. 
> If a tool -- especially a tool with the ubiquity and authority of Topcat -- 
> simply reports that the service is broken, and how[1], then the service 
> maintainers will be forced to fix their broken service, albeit at the cost of 
> some short-term irritation.

I take your point, but ...

> Postel's law is an engineering heuristic, not an architectural principle.

... a tool such as Topcat is a bit of engineering rather than an 
architectural component.

Probably the best thing for a user-directed application (rather than,
say, a validation service) to do is to work out how the data is really
encoded in the most reliable way (in this case magic numbers work well)
and log a warning if the MIME/HTTP tagging is inconsistent with this,
encouraging the user to contact the service about it.

-- 
Mark Taylor   Astronomical Programmer   Physics, Bristol University, UK
m.b.taylor at bris.ac.uk +44-117-928-8776 http://www.star.bris.ac.uk/~mbt/