requests for nALFS2

Kevin P. Fleming kpfleming at
Wed Aug 11 06:45:10 PDT 2004

Jamie Bennett wrote:

> will still have to be supplied. Just specify the filename and :-
> 	<unpack>
> <reference></reference>
> 		<archive>&source-dir;/gzip-1.3.5</archive>		
> 		<destination>&build-dir;</destination>
> 	</unpack>

Well, this is not a backwards compatible change, so it'd have to be 
documented well :-)

The problem I see with it is as long you are supporting downloads, you 
pretty much have to do digest checking. The download process can fail 
for so many reasons, and a lot of them don't get passed back into nALFS 
for it to know that the failure occurred. The digest check is the only 
way to ensure that the file present is what it is expected to be.

Let me suggest another alternative: how about some kind of wrapper 
element that would be used to specify that only _one_ (any one) of its 
child elements need to be successful. Sort of like {or}, except not 
conditional execution. This could be used to wrap _two_ (or more) 
{unpack} elements, one for each compression type that the package is 
distributed in. Each one would have all the information needed to 
process the package, and if one succeeded the profile could continue on 
ignoring the one(s) that failed. If they all failed then the profile 
execution would stop as it does now.

More information about the alfs-discuss mailing list