On 09/11/2015 06:06 PM, Alexander Shorin wrote:
> JWT is all good except one moment: it's not an alternative for OAuth (:

Yeah, that's true. However there are use cases where the same problem
can be solved by using JWT instead of OAuth, maybe even in a better way.

> And it's hard to say that it's simpler, especially in case of support
> of all the algorithms on browser side. WebCrypto is not a common thing
> yet.

I don't see a point in supporting all available algorithms. Assuming
that we restrict ourself to only implementing the HS256 and RS256
algorithms (at least as a start), which I think would be reasonable
choice, it's likely going to be simpler both from a implementation and
user perspective. Both HS256 and RS256 can be easily used in browsers.
There's even a browser-based debugger available at http://jwt.io/.

> But I'm +1 for JWT support in anyway. It has own good use cases.

Great :)

> P.S. Basically, CouchDB cookies are JWTs, except that payload isn't
> JSON, but binary Erlang term.

Yeah, they're quite similar to HS256 JWTs :)

One limitation with cookies that I've recently been facing is that you
can't rely on them as soon as requests to CouchDB are cross-origin, as
browsers often don't accept third-party cookies. This could be addressed
by including a JWT in the response to requests against the /_session
endpoint, which then could be used for authentication instead of a cookie.

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to