[Ndn-interest] Freshness & Latest Content

[Urs Schnurrenberger]

Hi Nick,

agreed. Thank you for the fingerpost that seems worth following. I first have to dig a little bit deeper and may get back to you later.

Best regards,
Urs

Von: Nick Briggs <nicholas.h.briggs at gmail.com>
Gesendet: Freitag, 4. Mai 2018 01:06
An: Urs Schnurrenberger <urs.schnurrenberger at unibas.ch>
Cc: Junxiao Shi <shijunxiao at email.arizona.edu>; ndn-interest at lists.cs.ucla.edu
Betreff: Re: [Ndn-interest] Freshness & Latest Content


On May 3, 2018, at 3:33 PM, Urs Schnurrenberger <urs.schnurrenberger at unibas.ch<mailto:urs.schnurrenberger at unibas.ch>> wrote:

Hi Junxiao,

thanks for the answers and the link. I have a follow-up question:

What then happens when the (same as below) Interest
        /nytimes/frontpage | CanBePrefix=true | MustBeFresh=true
arrives at a cache with the sole content
        /nytimes/frontpage/addBanner/v23 | 60'000ms
?

According to prefix matching and a positive freshnessPeriod, will I receive the addBanner only and never the complete front page?

Yes, but the publisher would be shooting themselves in the foot to publish it with such a structure.  They're much more likely to publish

                /nytimes/frontpage/v105/adBanner/v23 | 60000 ms

They would also be smarter to publish with a date coded version so an application could make an educated guess at the version it should be retrieving.
Namespace design for an application (or class of applications) is really important in order for things to work smoothly.

This issue was why the original CCN protocol specification had  MinSuffixComponents/MaxSuffixComponents optionally on the Interest so the application wouldn't get shut out by longer names that were long lived and shadowed the shorter name you were looking for.   I think that got dropped somewhere in the evolution of NDN, but I don't know where.

Best regards,
Urs

Von: Junxiao Shi <shijunxiao at email.arizona.edu<mailto:shijunxiao at email.arizona.edu>>
Gesendet: Donnerstag, 3. Mai 2018 23:37
An: Urs Schnurrenberger <urs.schnurrenberger at unibas.ch<mailto:urs.schnurrenberger at unibas.ch>>
Cc: ndn-interest at lists.cs.ucla.edu<mailto:ndn-interest at lists.cs.ucla.edu>
Betreff: Re: [Ndn-interest] Freshness & Latest Content

Hi Urs

I have some conceptual questions about freshness and how to ask for latest content.

Imagine 2 properly versioned versions of /nytimes/frontpage (properly in the sense of "larger" version = "newer" version, according to [1]):
format : contentName | freshnessPeriod
/nytimes/frontpage/v100 | 100ms [stored at cache X somewhere between requester and producer]
/nytimes/frontpage/v105 | 300ms [stored at producer]

I assume that if a user does only know about the contentName (prefix), but not about the latest versionNumber, he/she asks for
Interest: /nytimes/frontpage | CanBePrefix=true | MustBeFresh=true

Yes, that's what consumer should ask for.

Question 1:
Is it correct that this Interest is satisfied by cache X because there is a prefix match on the contentName and a freshnessPeriod > 0?

Yes, X will return /nytimes/frontpage/v100 until it becomes non-fresh.

Question 2:
Assuming the answer to Q1 is 'yes' and that the reuqester does NOT know the latest versionNumber, is there really no possibility for the requester to get the globally latest version v105 directly from the producer?

It is possible but requires changes on producer. See https://redmine.named-data.net/issues/4396#note-1


Question 3:
Assuming that the requester KNOWS the latest versionNumber (from some miraculous oracle), I guess she/he can simply ask for (/nytimes/frontpage/v105 | CanBePrefix=false | MustBeFresh=true). Correct?

Yes, the Interest should have version number if the consumer knows it.


Yours, Junxiao
_______________________________________________
Ndn-interest mailing list
Ndn-interest at lists.cs.ucla.edu<mailto:Ndn-interest at lists.cs.ucla.edu>
http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndn-interest/attachments/20180504/0f7ad877/attachment-0001.html>