Re: [ppsp] Call for WG consensus on Tracker Protocol encoding

Hi all

I support Joao’s, opinion on text based messages.

In ITU-T there is a long record of binary encoding due to bandwidth constrains but in IETF there is a large support for text based, due to its higher readability and simplicity for debugging.

Concerning the syntax of the messages I see at least three alternatives:

- SDP 

- XML

- JSON

SDP is very successful e.g. for SIP but is not used for new protocols for some time now

Between XML and JSONB I definitely support JSON for a number of reasons,  namely because it is ligher and more readable than XML, and has more support in IETF, as you can read in this article:

 <http://www.internetsociety.org/articles/using-json-ietf-protocols> http://www.internetsociety.org/articles/using-json-ietf-protocols

Best regard

Mário Nunes

From: ppsp [mailto:ppsp-bounces@ietf.org] On Behalf Of João Pedro Taveira
Sent: 13 de novembro de 2014 13:14
To: Huangyihong (Rachel); Zongning; ppsp@ietf.org
Subject: Re: [ppsp] Call for WG consensus on Tracker Protocol encoding

Hi to all,

About the specification language to used, I don't prefer this or other language. Arno already sugested XDR and Rachel sugested JSON or YANG. I think that base protocol is well defined in high level point of view. The main issue is on-the-wired encoding. 

Personally, I prefer binary encoded protocols. They are compact, and if well defined, they are simple and very efficient. The problem is that they are complex to use and/or implement.

The Tracker Protocol should be text encoded (imho, JSON) on-the-wired over HTTP, for now. This will allow faster testing and adoption. Since messages are historically defined in a hierarchical way from XML sharing elements, we will be able to discuss binaries hierarchical structures (e.g. mp4box or something like icmp) to get a binary encoding that match current abstract protocol.

For now, I think we should use JSON over HTTP.

Best Regards,

João Pedro Taveira

On Thu Nov 13 2014 at 8:05:21 AM Huangyihong (Rachel) <rachel.huang@huawei.com <mailto:rachel.huang@huawei.com> > wrote:

Hi  all,

IMHO, text encoding makes more sense to me. It’s prevalently used in application protocols over HTTP (e.g., alto protocol). And maybe we could use standard languages, like JSON or YANG to specify our messages (not just current C language). As for the efficiency issue,  it may be more efficient when parsing the binary encoding than text encoding. But the advantage is very limited from my point of view. And I don’t see such  efficient issue raised in current web servers (which accept text encoding). 

BR,

Rachel

From: ppsp [mailto:ppsp-bounces@ietf.org <mailto:ppsp-bounces@ietf.org> ] On Behalf Of Zongning
Sent: Thursday, November 13, 2014 10:28 AM
To: ppsp@ietf.org <mailto:ppsp@ietf.org> 
Subject: [ppsp] Call for WG consensus on Tracker Protocol encoding

Hi, all,

Firstly, thanks to the co-authors of Base Tracker Protocol for persistently moving the draft forward.

Now the outstanding issue is that as a Standard Track document, we really NEED to agree on a mandatory encoding for the interoperable on-the-wire Tracker Protocol. As discussed in the PPSP session today, we will start WGLC for the draft, provided that we can reach a rough consensus on the encoding option and the co-authors revise the draft accordingly.

Currently we have two options mentioned in the draft – they are text based and binary based. For an exemplary comparison, please see Section 3.1 in the draft. Could folks in the group give their opinions on which encoding option is preferred for Tracker Protocol, and why? People are welcome to give other options beyond text and binary, but please do show us the reason for choosing them.

The during of this call will not last too long, before the co-chairs will make a decision. So, please do contribute your technical expertise in this perspective, to enlight the group.

Thanks.

-Yunfei & Ning

_______________________________________________
ppsp mailing list
ppsp@ietf.org <mailto:ppsp@ietf.org> 
https://www.ietf.org/mailman/listinfo/ppsp