Anthony Bryan wrote:
> 
> I've seen this come up a few times, and since we're changing things...
> 
> maybe we should rename "preference" to "priority" or something else
> that's more intuitive.
> 
> and change the numbering scheme while we're at it, if need be.
> 
> any suggestions?
> 

Something that definitely needs clarifying:

Is it an absolute priority/preference? Should clients *always* try
higher-priority mirrors and only if they fail continue with the lower
priority mirrors? Or does it change the probability of being selected in a
random choice of mirror?

Example: if I have mirror A with preference 10 and mirror B with pref 20
(assuming higher number is higher preference), does that mean mirror B is
twice as likely to be used, or does it mean mirror A will be used if and
only if mirror B can't be used?

Or... should there be a way for a metalink author to choose between both
behaviors?

RFC 3782 gives a way to provide *both* a priority and a weight (see page 2
and 3). Equal-priority choices are randomly chosen, with probability of
selection proportional to the given weight. I think this looks quite
useful, maybe we could borrow it. A metalink could give weight depending on
available bandwidth, and another could give a main server and another
mirror that is strictly for fallback.



--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"Metalink Discussion" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/metalink-discussion?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to