- From: Ryan Sleevi <sleevi@google.com>
- Date: Mon, 5 May 2014 07:31:29 -0700
- To: Rich Salz <rsalz@akamai.com>
- Cc: public-webcrypto-comments@w3.org, GALINDO Virginie <Virginie.GALINDO@gemalto.com>
- Message-ID: <CACvaWvbv1ECMH9sOC9iNEXc5Nw12U=yn-uJQ5pYOGz7YzOoaaA@mail.gmail.com>
On May 5, 2014 7:26 AM, "Salz, Rich" <rsalz@akamai.com> wrote: > > Ø As has been discussed - repeatedly - you can't programatically separate the algorithms into two (or more) namespaces, because once shipped, you can *never* migrate between them, as such migrations are inherently breaking API changes. > > First of all, it’s a one-way path. Once something becomes broken or weak, it never moves out of that category. > > > > Second, the suggestion is to make second-class citizens, from the very beginning out of those things that we know should be second-class. Since you are not (yet?) able to provide guidance on what to avoid, then make the names used in the code provide that guidance. JS has exceptions and when bad crypto is no longer supported as a first-class part of the API, then code that needs to use them can do catch the exception and try the other namespace. Defining that namespace in your base document will increase interoperability in the face of a changing crypto world. > > That is still a fundamentally broken API. All it does is result in the exact same behaviour you seem to think it will avoid - every developer will follow that exact same pattern for every algorithm. Those that don't, it is still a breaking change for the web - since example.com was working fine with My Awesome Browser v23, but now suddenly stops working in My Awesome Browser v24, because its complaining about uncaught exceptions. Practically speaking, your proposal is worthless for accomplishing its goals of security. The group did indeed discuss this at length, and thoroughly established that it provides no benefits. > > As for a UA asking for additional confirmation, what might that look like? “The createSignedBlob applet is asking to use AES in CBC mode, which is known to be weak; see http://iacr.org/preprints/.... For details. Proceed or Cancel.” Really? > Strawmen are very easy to argue against, but neither productive nor constructive towards doing so. > > > /r$ > > > > -- > > Principal Security Engineer > > Akamai Technologies, Cambridge, MA > > IM: rsalz@jabber.me; Twitter: RichSalz > >
Received on Monday, 5 May 2014 14:31:56 UTC