[AVT] Re: IESG Review of draft-ietf-avt-mpeg4-simple-07.txt - Discuss Comments

[jan.vandermeer@philips.com]

Dear Allison, all,

Thanks for the IESG comments. In response I would like to suggest the following:

1) Comment:  It is strange to have more than one section labeled
  "Introduction."  Please pick a new label for section 2.1.

My proposal is to label 2.1 as "Signaling by MIME format parameters"

2) Where is the security model defined for ECMAScript in this context?  
   (Problems with the model have been part of the Javascript security 
   problem for Web browsers.)

MPEG-4 defines some important constraints on the use of ECMA scripts in MPEG-4. An annex to the MPEG-4 system spec describes the differences between "regular" ECMA scipts and "MPEG-4 scripts". As far as I understand these constraints do not allow dangerous ECMA script constructs, which means there is no need for an ECMA script security model in this context. Below I attached the MPEG-4 annex that describes the differences. 
I suggest to resolve this as follows:
a) replace in section 5 "security considerations" all current references to ECMAScript by MPEG-4 script, and 
b) add to the fifth paragraph (starting with "In ISO/IEC 14496-1 a security model is defined for ...") the following trailing sentence "Note: MPEG-4 scripts are based on ECMA scripts, but there is no need for an ECMA script security model, as the use of insecure ECMA  script constructs is impossible in MPEG-4 scripts."

3) 2.4:  Why is this form of application-level fragmentation better than
   IP fragmentation?

This is because of error resilience. If IP fragmentation occurs without application-level fragmentation, then all data of the entire Access Unit gets lost when one such fragmented IP packet is lost. When application-level fragmentation is used, and one RTP packet with an AU fragment gets lost, then the received RTP packet(s)  with the other AU fragments can still be decoded. I suggest to remove "so as to avoid IP layer fragmentation" from the first sentence of section 2.4, and to add the following after the first sentence: "Hence when an IP packet is lost after IP-level fragmentation, only an AU fragment may get lost instead of the entire AU".

Any comments very welcome.


Best regards,

Jan van der Meer


*******begin of MPEG-4 system annex*******

        MPEG-4 Scripts Have a Rigid Representation  
MPEG-4 scripts differ slightly from ECMA scripts. The most important difference is that MPEG-4 scripts are not represented textually, but are transmitted as a parse tree representation.  This means that only constructs that can be represented by the MPEG-4 parse grammar can be encoded and transmitted. Not all ECMA script constructs can be represented in MPEG-4 scripts.

The differences between ECMA scripts and scripts that can be represented in MPEG-4 are given below.

        Keywords
MPEG-4 scripts cannot utilize the following keywords: catch delete do finally in instanceof throw try typeof void with .

This means that  do – while loops and  for – in loops are not possible.

        Relational operators

The relational operators  "===" and "!==" cannot be included in MPEG-4 Scripts.

        Labeled statements

In MPEG-4 scripts it is impossible to label statements and to break or continue to labeled statements.

        Switch statement restriction

MPEG-4 scripts with switch statements can only take numerical case expressions and always must have at least one case statement.

In particular this means that

switch {

        case (x+1):  ….

}

is not possible, while

Switch {

        case 1:  ….

}

is okay.

        Functions, not programs

The MPEG-4 event driven script model allows only functions to be called in response to events.

        Expressions

Statements that include statement blocks, such as for are represented in the parse tree as having an empty statement block, where as in ECMA script they can omit this block. Functionally, the statements behave identically. For example, the expression:

        for ( <expr>; <expr>; <expr>)

must be represented as

        for ( <expr>; <expr>; <expr>) {}

        Array and Object Literals
Array and object literals of the form [value1, value2, .., valueN] and {property1:value1, property2: value2, .. propertyN:valueN} cannot be used in MPEG-4 scripts.

*******end of MPEG-4 system annex*******

Allison Mankin <mankin@psg.com> 2003-06-27 06:19 AM Please respond to mankin

To:        Jan vanderMeer/EHV/CE/PHILIPS@EMEA3 dmackie@apple.com viswanathan.swaminathan@sun.com singer@apple.com Philippe Gentric/MP4-SUR/CE/PHILIPS@EMEA1         cc:        casner@acm.org magnus.westerlund@ericsson.com csp@csperkins.org smb@research.att.com ned.freed@mrochek.com mankin@psg.com avt@ietf.org         Subject:        IESG Review of draft-ietf-avt-mpeg4-simple-07.txt - Discuss Comments         Classification:

The IESG reviewed draft-ietf-avt-mpeg4-simple-07.txt and had a few concerns
that should be addressed before the draft can advance.  Editorially:

   Comment:  It is strange to have more than one section labeled
  "Introduction."  Please pick a new label for section 2.1.

Steve Bellovin and Ned Freed both request that a reference be given in
the Security Considerations for security model for ECMAscript.  Here
is Steve's Discuss comment:

   Where is the security model defined for ECMAScript in this context?  
   (Problems with the model have been part of the Javascript security 
   problem for Web browsers.)

Steve also asked:

   2.4:  Why is this form of application-level fragmentation better than
   IP fragmentation?

Please discuss the issues in email and we'll see if the fixes to the draft
can be done as notes rather than a revised i-d, to be quicker.

Allison