OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help

office message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]

Subject: Re: [office] Re: encryption

2009/9/1 Robert Weir/Cambridge/IBM <robert_weir@us.ibm.com>
The reason to allow more than one algorithm, aside from preferences
(individual, corporate, national requirements etc.) is that an attack
could be found against any one of these algorithms and you don't want to
be in a situation where the only algorithms specified are weak or broken.
The use of SHA1, in particular, does not seem to be a good algorithm

Agreed.  Being bound to SHA1 is not sensible.  And if the situation were to be that all the algorithms we specified are somehow compromised, then yes, we end up with egg on our faces if we have no way of allowing the new stronger algorithm.

But to be honest, I really don't think that's a fair reflection of the situation.

If you look at actual public sector requirements (eg. the South African MIOS, the Brazilian Ping, European EIF etc) I think the usual practice is to specify a list of acceptable algorithms and to update and maintain that list from time to time.  I haven't seen any example of allowances for additional as yet undefined algorithms.  So long as ODF specifies a reasonable subset of those algorithms (not just one) I believe we effectively meet the requirements.

To hold open a place holder for the as yet unnamed Uber-algorithm which is just around the corner and which will displace the existing ones might seem on the surface to make sense and that is the argument which is frequently used.  But what may be more likely is the mechanism being used in the exact opposite way ie. the space will be used by implementation defined (and weaker), "legacy" algorithms.  We have already seen the real difficulties raised by supporting the legacy hashing algorithm in OOXML.

Of course, this doesn't mean you need to leave it open ended.

It really boils down to three questions:

1) For each algorithm type (hash, encryption, etc.), what unique
identifier to we associate with each algorithm?

2) For the sake of encouraging interoperability do we recommend or mandate
that a subset of these algorithms be supported?

3) Do we allow implementation-defined algorithms beyond those which we
have assigned identifiers to?

I am happy that we do, BUT that would fall into the category of an extension which implies a different conformance level.  Either that or disallow it.


But remember, there is nothing in the standard that mandates the support
of the document encryption feature at all, so #2 doesn't really help us
much here, does it?


Bob Jolliffe <bobjolliffe@gmail.com>
Ming Fei Jia <jiamingf@cn.ibm.com>
09/01/2009 12:39 PM
Re: [office] Re: encryption

Thanks Ming Fei.  You summarized my concerns much better than I ..

What was the original intent in specifying SHA1 and Blowfish?  It seems to
me, though I wasn't around at the time, that the idea was primarily to
ensure interoperability, perhaps above other plausible goals.  The
selection of a widely available public domain cipher seems to reinforce
that interpretation.

The casualty of interoperability here was choice.  There was no choice.
By allowing documented algorithms (as per xmlenc-core) we open the window
of choice slightly whilst maintaining some hope of interoperability.  This
seems like it might be a good thing.

By opening up the third option (implementation defined algorithms) we
maximize the choice but, as Ming Fei says, we risk the standard having no
meaning or relevance regarding encryption.  This might be reasonable
tradeoff under some conditions.  In the case of hashed passwords there is
use case of conversion of legacy documents (which I'm still not that
comfiortable with).  In the case of encrypted XML streams I don't think
the same argument applies.

Is this perhaps yet another case for discrimination on the grounds of
conformance class, where the use of an implementation defined algorithm is
not disbarred, but it is treated essentially as an extension conforming to
a different, less strict, class of document?


2009/9/1 Ming Fei Jia <jiamingf@cn.ibm.com>
In the proposal:
The defined value for the "algorithm" attribute is 3 options:
? The Blowfish algorithm in CFB mode.
? An IRI defined in §5.2 or §5.3 of [xmlenc-core]: The algorithm specified
in §5.2 or §5.3 of [xmlenc-core] for this IRI.
? An IRI specifying an implementation defined algorithm.
Actually I think the proposal means ODF has no restriction for encryption
algorithm and ODF encryption algorithm could be anything. Then, does the
standard have meaning here? Of course, that is OK if there is some
exception that everyone believes:
(1)Encryption algorithm does not have any interoperability issue in
(2)Encryption algorithm will have no interoperability issue in the future
(3)Implementation defined algorithm is not conforming to ODF
(4)Standard here can not solve problems at all even the issues are there.
or anything else?

We could have some trade-off for the real complexity, but I suggest to be
careful to evaluate this extending. thanks a lot.
Best Regards,

Mingfei Jia(???)
IBM Lotus Symphony Development
IBM China Software Development LAB, Beijing
Tel: 86-10-82452493 Fax: 86-10-82452887
NOTES:Ming Fei Jia/China/IBM E-mail: jiamingf@cn.ibm.com
Address: No.28 Building, Zhong Guan Cun Software Park, No.8 Dong Bei Wang
West Road, ShangDi, Haidian District, Beijing 100193, P.R.China

Bob Jolliffe ---2009-08-31 22:46:31---In addition - I don't know the
answer to this, but in the interest of uniformity, is there also an IRI
which can used to indica


Bob Jolliffe <bobjolliffe@gmail.com>




2009-08-31 22:46


[office] Re: encryption

In addition - I don't know the answer to this, but in the interest of
uniformity, is there also an IRI which can used to indicate blowfish?
Then we are clear the value of the attribute is an IRI.

2009/8/31 Bob Jolliffe <bobjolliffe@gmail.com>
what I was trying to say on the call is that we now have 3 options for
each algorithm, including a catchall "implementation defined" IRI.  I
would prefer to see this last option allowed but not recommended.


To unsubscribe from this mail list, you must leave the OASIS TC that
generates this mail.  Follow this link to all your TCs in OASIS at:

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]