Re: More namespacing woe ...

[Michael Meeks <michael helixcode com>]

On 30 Nov 2000, Miguel de Icaza wrote:
> Mhm, you raise an interesting point.  From the moniker perspective,
> using #gzip: should compress the parent, and return data compressed on
> the other end.
  
        The #gzip: moniker is a moniker that maps a Stream <-> another    
Stream, the compression semantics I would have thought are transparently  
obvious. If A is the parent Stream and B is the gzip stream then if you   
read from B you get A unzipped and if you write to B you write zipped data
to A hence:

        A       <-> gzipping process moniker    <->     User

        I see no reason for confusion here, the only problem comes when   
people want to read and write at the same time, this presents a rather    
more difficult problem; however, as a first pass we can select which
direction based on the first read / write. But either way the moniker is  
going to be fun to write to take account of seeking etc. So; I think we   
just bite the bullet, realize it is a hard moniker to write and spend some
time getting it right.

        Eg. it would be easy to have a cache of the uncompressed data on  
disk / in memory that would be sent to the parent stream as the gzip proxy
stream's reference dropped to 0.

> So I think we should keep #gunzip: for gunziping, and #gzip for
> gzipping.
  
        Worryingly I have to agree with Maciej; this not a good idea.       

        Regards,
        
                Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot