Fetching

[Jeremy Lavergne]

An anonymous user appeared on IRC today to discuss segmented  
downloading.  While he had suggested we allow the user to pick a  
downloading client, we pressed our goal of staying with as much system- 
provided utilities for core operations as possible.

To that end, I'd like to start up some discussion in our mailing list  
regarding a few aspects of the fetch phase.

With the advent of 1.8.0, we will have a faster dependency search.   
This will allow us to quickly find all the ports associated with a  
build and begin downloading them.  While we could perform all these  
downloads before we commence building, why not have them be done in  
parallel [1]?

Similarly, taking the aspect of segmented downloading, would we be  
interested in having an even number of segments as there are mirrors  
for a port?  That is, one instance of curl will download the first  
third of the file from the main site, a second third from the first  
mirror, and the final portion from the last mirror; hopefully all  
three finish in the same time as a third of the original download  
since you may be maximizing your download capacity [2].

I feel that the latter functionality should be an opt-in feature, as  
some users may not want to suck up their bandwidth. The former should  
definitely be implemented as there is no reason to wait for a build to  
finish before fetching more files.

Your thoughts, gentlemen?

  [1] http://trac.macports.org/ticket/2421
  [2] https://trac.macports.org/ticket/20652

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2435 bytes
Desc: not available
URL: <http://lists.macosforge.org/pipermail/macports-dev/attachments/20090812/ae20d0b7/attachment.bin>