Change to bzip2?

[Jeff Johnson]

Steve G wrote:

>>Anyways, getting headers compressed (which are typically 10-15% of package 
>>size), would presumably save 5-7.5% or so of package size, probably a bigger 
>>saving than anything else yet proposed.
>>    
>>
>
>I like this idea, too. Would there be impacts to rpm querries?
>  
>

Zero, I speak of headers in packages, not headers in the database, where 
speed
of retrieval, not size, is clearly the operant metric.

The legacy issues are show stoppers, however, and voiding header+payload 
signatures
(which need to be abandoned for lots of reasons, 2 signatures schemes is 
worse
than either) will put the kibosh on any change afaik.

Oh well ...

FWIW:

Headers are gonna change in rpm-4.4.2 to permit use of xml to represent
header content, and there's four or five other flaws that need addressing
at the same time (like permitting an array of binary elements, and adding
an ASN.1 like type to ease tertirary lookup of data embedded in tags
(think: signature/pubkey fingerprints.)

If I gotta break something in rpm package format, I might as well break 
everything,
the degree of pain is exactly the same imho.

73 de Jeff