I realize it's late to make changes, but I really think unifying type &
mode makes a lot of sense. Here's my suggestion: extend the "resource
directory" to include a generalized description of the cost-types that the
server provides. I won't suggest a syntax -- too early for that!! -- but
here's what the description should provide for each distinct type:
* The type name. This is what clients give in the cost-type parameter.
* The value type. "numerical" & "string" are obvious possibilities.
* Linear: a boolean flag. True means this numeric value is linear.
False means values are not linear. This takes the place of
the ordinal/numerical mode.
* Category: What this cost measures: "delay", "hops", "cost", "other".
Here "cost" means money.
* Description: An optional free-form description of this cost type.
For human consumption, to be used when investigating a new ALTO server.
* URL: An optional url with a detailed description of this cost type.
For more involved "Descriptions", of course. This can be in
multiple languages, of course.
Yes, the change would be painful, but I think it will simplify the
protocol, servers and clients -- AND it will make it easier for servers to
add new cost types.
To simplify clients, we might reserve a few type names, such as
routingcost-numerical
routingcost-ordinal
hopcount-numerical
hopcount-ordinal
and require them to have the obvious descriptions.
- Wendy Roome
On 01/29/2013 15:00, "[email protected]" <[email protected]> wrote:
>Date: Mon, 28 Jan 2013 23:08:23 +0000
>From: "Reinaldo Penno (repenno)" <[email protected]>
>To: "[email protected]" <[email protected]>
>Subject: [alto] ALTO Protocol Outstanding Issue II: Unifying cost-mode
> and cost-type to a single type
>Message-ID:
> <[email protected]>
>Content-Type: text/plain; charset="us-ascii"
>
>Hello,
>
>Please send your comments to the list in the next 2 weeks. Silence means
>we will go with the proposal.
>
>Discussion II: Unifying cost-mode and cost-type to a single type
>
>e.g., routingcost-num and routingcost-ord
>
>Having a single type simples the protocol since there is just one
>parameter when indicating cost. But it will impact current
>implementations and might loose flexibility.
>
>Proposal: Leave it as is.
>
>Thanks,
>
>Reinaldo
>
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
