Wendy,
Thanks for your great work! Answering the question from
Richard, Page 4: Can a sever provide both a numerical and an ordinal cost map?
According to RFC7285, an ALTO sever MUST support at least one of the
following modes: numerical and ordinal. Which should means, an ALTO sever can
provide both a numerical and an ordinal cost map in some cases. In this case,
no matter the client desires a numerical or an ordinal mode, the sever can meet
its demand.
Best,
Haoran
_____________________________
From: "Y. Richard Yang" <[email protected]>
Sent: 星期六, 六月 6, 2015 12:39 上午
Subject: Re: [alto] ALTO Interop test in July
To: Wendy Roome <[email protected]>
Cc: "[email protected]" <[email protected]>
Wendy,
Great work! Here are some early comments.
- Page 4: "If a server provides a numerical-mode cost map, ..." =>
"If a server provides a numerical-mode cost map for the "routingcost" metric.,"
The reason is to make clear that Figure 2 is for the only routingcost metric
only.
- Page 4: "If a server provides an ordinal-mode cost map, the server
may use whatever values it wants, provided the ordinal values preserve the
order of the numerical values. "
I see that there are duplicate values in Figure 2 (e.g., 75.0). A
question is to define the meaning of "preserve", as RFC7285 does not define it,
I believe. Here is an attempt in defining it precisely. First, for non-equal
case, we should have num[x] > num[y] => ord[x] < ord[y]. An issue is if we
define the case for num equal, e.g., num[x] == num[y] => ord[x] == ord[y]. I
assume that we have it?
- Page 4: Can a server provide both a numerical and an ordinal cost
map?
- Page 5: "a numerical-mode cost map MUST use these values, and an
ordinal-mode cost map may use any values consistent with this ordering."
Page 4 used the word preserve and here it is consistent. How about the draft
uses one word, say consistent, and the draft defines the meaning consistent as
above?
Similar small consistent use for the alternate map.
- When I read Page 7: "A server MAY provide whatever additional
resources it desires, as long as they are consistent with the network maps,
cost maps and endpoint properties defined in Section 2. ", a question of 3
network maps came to mind. I liked that you clarified on page 8.
- Page 7: "Cost Map resources for the "routingcost" and/or "hopcount"
metrics, in either "numerical" or "ordinal" modes, using the values defined
above." How about clarify that it is for the alternate network map?
- Page 9: "However, each client MUST be able to verify the default
network map and the associated "routingcost" cost map." Does this handle the
case of both modes, if provided?
Overall, a great design, and I am looking forward to seeing the ECS
test cases as well :-)
Richard
On Thu, Jun 4, 2015 at 3:34 PM, Wendy Roome
<[email protected]> wrote:
To get the ball rolling on an ALTO interop test, I have submitted
draft-roome-interop-ietf93-00. Please, please, someone else please read it.
https://urldefense.proofpoint.com/v2/url?u=https-3A__datatracker.ietf.org_doc_draft-2Droome-2Dinterop-2Dietf93_&d=AwICAg&c=-dg2m7zWuuDZ0MUcV7Sdqw&r=4G36iiEVb2m_v-0RnP2gx9KZJjYQgfvrOCE3789JGIA&m=bqET3V84WZLd20R9K89VK4Hm0aQzcWdZF7f9JgZZ42Q&s=xMKyCEnwH7PvUBYG16p4mpynr8QXYjHQakn4U8R_mK4&e=
It specifies quite a lot: two network maps, two cost metrics, and a custom
property. However, only *required* resources are the default network map,
one routingcost cost map (numerical or ordinal), and an EPS for the
default network's pid property. Everything else is optional.
- Wendy Roome
_______________________________________________
alto mailing list
[email protected]
https://urldefense.proofpoint.com/v2/url?u=https-3A__www.ietf.org_mailman_listinfo_alto&d=AwICAg&c=-dg2m7zWuuDZ0MUcV7Sdqw&r=4G36iiEVb2m_v-0RnP2gx9KZJjYQgfvrOCE3789JGIA&m=bqET3V84WZLd20R9K89VK4Hm0aQzcWdZF7f9JgZZ42Q&s=0CuB6ojyak7XxLmBQ2liCZzG5JiQaSfxqSPnBloJLX4&e=_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
