[apps-discuss] Review of draft-ietf-alto-protocol-11.txt

Ted Hardie <ted.ietf@gmail.com> Thu, 26 April 2012 21:40 UTC

Return-Path: <ted.ietf@gmail.com>
X-Original-To: apps-discuss@ietfa.amsl.com
Delivered-To: apps-discuss@ietfa.amsl.com
Received: from localhost (localhost []) by ietfa.amsl.com (Postfix) with ESMTP id 8444A11E809B for <apps-discuss@ietfa.amsl.com>; Thu, 26 Apr 2012 14:40:31 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -3.569
X-Spam-Status: No, score=-3.569 tagged_above=-999 required=5 tests=[AWL=0.030, BAYES_00=-2.599, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([]) by localhost (ietfa.amsl.com []) (amavisd-new, port 10024) with ESMTP id WU8gz2w5jNfH for <apps-discuss@ietfa.amsl.com>; Thu, 26 Apr 2012 14:40:30 -0700 (PDT)
Received: from mail-vx0-f172.google.com (mail-vx0-f172.google.com []) by ietfa.amsl.com (Postfix) with ESMTP id 0486821F883A for <apps-discuss@ietf.org>; Thu, 26 Apr 2012 14:40:29 -0700 (PDT)
Received: by vcbfo1 with SMTP id fo1so73850vcb.31 for <apps-discuss@ietf.org>; Thu, 26 Apr 2012 14:40:29 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:date:message-id:subject:from:to:content-type; bh=vOqYhJorrJWhHncrw26vOJE50g5SJ0HaAtkDr8y+UJE=; b=vTPSbdU1n4kU/ka8urOSO/ImeOe3H0MXxSn6eyvLkGli9TKxo18SAdT1M5ayDhOzoQ YuPjxv2d5QTTfJ/iBCf/1BfxRnzQykVkfwhgLN8Yor+NOLdEoY1SIqRqCbwJZ3QjmhN/ dTSek/bt/nK0H3GDoWEZQlCC+PD4779VLgfpAQIkuxZ9shipNXNkmYJhb20OUAc+PMVU fd7EItZD7F0S5bOcEQr4YbdND9S0mICTvi+i9Q8vdRxnBA8pFjsjNbuCHO8mVVL9pGGj m57e9PJ1t1TF/rEhQPJgK2DjNEiWVyg8LtiEFTbasVjGXM1sQi3A3M5qL+oflwpy1P3Z rt/A==
MIME-Version: 1.0
Received: by with SMTP id s15mr8743370vcv.60.1335476429508; Thu, 26 Apr 2012 14:40:29 -0700 (PDT)
Received: by with HTTP; Thu, 26 Apr 2012 14:40:29 -0700 (PDT)
Date: Thu, 26 Apr 2012 14:40:29 -0700
Message-ID: <CA+9kkMB4nUayYO3q8ydj477jh9NM8oZyXAQ5k-NsRN=KHjb3cA@mail.gmail.com>
From: Ted Hardie <ted.ietf@gmail.com>
To: draft-ietf-alto-protocol.all@tools.ietf.org, apps-discuss@ietf.org
Content-Type: text/plain; charset=ISO-8859-1
Subject: [apps-discuss] Review of draft-ietf-alto-protocol-11.txt
X-BeenThere: apps-discuss@ietf.org
X-Mailman-Version: 2.1.12
Precedence: list
List-Id: General discussion of application-layer protocols <apps-discuss.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/apps-discuss>, <mailto:apps-discuss-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/apps-discuss>
List-Post: <mailto:apps-discuss@ietf.org>
List-Help: <mailto:apps-discuss-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/apps-discuss>, <mailto:apps-discuss-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 26 Apr 2012 21:40:31 -0000

Please resolve these comments along with any other Last Call
comments you may receive.  Please wait for direction from your
document shepherd or AD before posting a new version of the draft.

Document: draft-ietf-alto-protocol-11.txt
Title: ALTO Protocol
Reviewer: Ted Hardie
Review Date: April 26, 2012

Summary: Despite the length of this review, the document is
fundamentally ready for publication on the standards track.  There
are sections which need clarification or re-consideration, but
they are not fundamental to the protocol's operation.


Section 5.1 needs clarification.  While it clearly introduces the
types for the cost type and cost mode parameters, it buries the type
for the cost itself within the description of the ordinal cost mode: Cost Mode: ordinal

   This Cost Mode is indicated by the string 'ordinal'.  This mode
   indicates that the costs values to a set of Destination Network
   Locations from a particular Source Network Location are a ranking,
   with lower values indicating a higher preference.  The values are
   non-negative integers.

This makes it unclear whether Costs of the numerical Cost Type are
similarly restricted to non-negative integers. Based on other
descriptions of ordinal comparison (e.g. q values in content
negotiation), I also presume that the clients are not expected to
infer anything about the weights based on the integers chosen for the
ordinal type. That is: 3 PIDS with costs of 2, 4, 6 should be treated
the same if the assigned values are 3, 17, 18, because the order is
the same.  Additional text to highlight this (or provide the
appropriate treatment) would be useful.  The section would also
benefit from an example of a numerical operation other than

In Section 7.2, the document describes the request/response traffic as
based on "standard HTTP"; the security considerations sections appears
to recommend TLS for integrity protection in 11.3 (this property is
not mentioned in 7.2.5, which lists TLS as MAY for authentication and
encryption).  This should be reconciled; if 11.3 is not recommending
TLS or any object-based security, more detail is probably needed on
the consequences of an attacker controlling this data.


General comment: there is a great deal of material in this document
that is either tutorial in nature or which defends a particular set of
design choices (e.g. section 6.1).  While this material is not
inherently bad, it is interleaved into the specification in ways that
make it difficult to extract the concrete protocol specification.  If
there were sections that the WG felt could be moved to an appendix
summarizing the benefits of the ALTO approach and its choices, the
spec might get significantly tighter and easier to read.

In Section 4.1, the concept of a PID as an identifier is introduced,
but without a pointer to the definition of the identifier type.  There
is a forward reference to section 5, but this describes cost maps, not
the PID identifier itself.  Reading (or grepping) forward provides a
pointer to PIDName, but without a succinct relationship defined.  I
suggest adapting the description in for this purpose:

"A PID is a US-ASCII string of type PIDName (see section 7.5.1) and
its associated set of endpoint addresses."

I believe this could be inserted between the first two sentences of
the first paragraph in section 4.1

In Section 4.2.1, the document says:

    A RECOMMENDED way to satisfy this property is to define a PID that
   contains the prefix for IPv4 or ::/0 (for IPv6).

This assumes that the cost map provided includes all IP addresses.
While this may be true for some use cases, there are other use cases
where it may not be true, and the text here should probably be
modified to handle those cases.  Something like

"A RECOMMENDED way to satisfy this property is to define a PID with
the shortest enclosing prefix of the addresses provided in the map.
For a map with full IPv4 reachability, this would mean including the prefix in a PID; for full IPv6 reachability, this would be
the ::/0 prefix"

might suit the case.

In section 7.2.2, the document says:

   "Note that it is possible for an ALTO Server to employ caching for the
   response to a POST request.  This can be accomplished by returning an
   HTTP 303 status code ("See Other") indicating to the ALTO Client that
   the resulting Cost Map is available via a GET request to an alternate
   URL (which may be cached)."

The phrase "employ caching" is ambiguous here, as the results of the
initial POST are not cacheable even in the case of a 303; only
the results of the later GET request are cachable.  Since cachability
is a major reason cited for the re-use of HTTP, some additional text
on the use cache control directives ("public" and "Max-age" seem
particularly important in this context) might also be useful.

In 7.4.2, the document describes the error resource, and notes this:

 reason  A (free-form) human-readable explanation of the particular

If I understand JASONString correctly, no language tag is supplied,
so human readability will be highly variable,  Fixing that, rather
than relying on a local mapping of the underlying code, is hard
work.  Dropping this optional aspect of the error resource may
be something for the WG to consider; this pushes localization
of the error code description to the endpoint.

The endpoint property resource seems to be specified in too many
sections for easy comprehension.  Section describes it, but
points to (the related capabilities) which in turn points to
3.1.3, which says:

   This service allows ALTO Clients to look up properties for individual
   endpoints.  An example endpoint property is its Network Location (its
   grouping defined by the ALTO Server) or connectivity type (e.g.,
   ADSL, Cable, or FTTH).

It is only in Section 10.3, with the creation of the registry, that
it is clear that these are registered tokens, rather than free text.
I suggest that this be clarified early, with a forward pointer to
the actual registry.

In section 9.3, the document notes that the ALTO client SHOULD use
STUN to determine a public IP for a source address.  However,
the description of the Endpoint Cost Service contains this text:

      If the list of Source
      Endpoints is empty (or not included), the ALTO Server MUST
      interpret it as if it contained the Endpoint Address corresponding
      to the client IP address from the incoming connection

It would seem a valid recommendation here would be to propose that the
CLIENT omit the Source Endpoint where it currently recommends the use
of STUN.  I infer that this is not recommended because the Endpoint
Address service may be provided within a NAT boundary shared with the
CLIENT, thus making the source address of the packet different from
that which would be seen externally (resolving this would require the
ALTO service to have general knowledge of the NAT bindings).  But the
opposite case is not handled.  What happens when the destination
address is within the NAT boundary of the client?  Won't the external
address determined by STUN will have a similar mismatch to the true
origin address for packets destined to the internal destination?

In 9.4, it may be useful to note that there are private ASNs which
are not unique, and that a single looking glass may therefore give
an incomplete view of the full mapping of IP addresses to ASNs.  In
general, this section seems to point to future work, and I suggest
that it more clearly say that a standardized method for this mapping
is left to future work.

In section 10.4, it may be useful to advise the Expert Reviewer
on how to handle registrations which differ only as to case;
for example, would PRIV: or EXP: be permitted in a registration?


The Abstract says:

   These maps provide simplified,
   yet enough information of a network for applications to effectively

This seems to be missing a noun or to require restructuring.  Possibly:

"These maps provide simplified information, but it is enough for
applications to effectively use."?

"Additional services are built on top the maps."  should be "on top of
the maps".

In the Introduction, "see below" could be shifted to a more precise
forward reference to a specific section.

In section 2.2, the document says:

   In particular, an ALTO Server defines network
   Endpoints (and aggregations thereof) and generic costs amongst them,
   both from the network region's own perspective.

I personally found this hard to parse.  Some re-wording may be useful.
(Possibly eliminating the ", both"?)

In Figure 1, the top most enclosing box is labelled "ISP".  While that is
certainly a common use case, there seem to be others. Given the
preceding text, would it be more general to use the term "Network

In section 3, the document says:

"Note that functionality offered in different
   services are not totally non-overlapping (e.g., the Map Service and
   Map Filtering Service)."

Would it be equivalent to say "functionality offered in different
services may overlap"?

Section 3 is titled "Protocol Structure", but the subsections are about
specific services, and the base text doesn't really describe what
most IETF readers will understand as a description of the structure of
the protocol (that seems to be in Section 7).  A different title (maybe
"Service framework"?) would seem to be a better fit.

I believe Section 4 might read more naturally if the second paragraph
were moved up to be first.

In 7.5.1, the encoding of PID Name is given in relation to a permitted
US-ASCII code range, except for the "." separator.  Providing the
related code point would be useful.  Several other sections use
"alphanumeric" as a limitation without giving the code point ranges;
in those cases, the actual code point ranges would be useful.

The Content-Length in some examples (e.g. in 7.6.3) are listed
as [TODO]; this should be fixed.

In section 9.2, the document says "one path my have a NAT64 middlebox)",
where my should be "may".

In Section 10.1, it is common for the change controller to be the IESG,
rather than the individual editors of the draft.