> > Yes. Otherwise you could subvert other users' result sets as
> > you don't know who created it.

> Not so. In SRW (unlike Z39.50) the result set name is really a result
> set identifier generated by the server rather than requested by the
> client. So in SRW the result set name effectively acts as a session id.

Yes. But if they persist, which they must in some form, then they can be
operated on.

For example, I send to client A a result set named 'rs1'. The rogue DDOS
client then sends me a request against a result set named 'rs1' which
promptly disappears for the real user.

In the time between the server sending the resultset name to the client, a
different attacking client can send a request which uses that name. You
simply can't avoid that. You need to have a way of determining if the
client is allowed to operate on that result set.

> This result set name only has limited life. One receipt of a second SRW
> request to get the next 10 records, the server is perfectly at liberty
> to respond with a new result set name (at an abstraction level this name
> would be referencing the same result set) i.e. this is just a mechanism
> to maintain state between SRW requests.

I could send continuous (SOAP is HTTP/1.1 so includes pipelining and
gzipping, making this even more effective) requests to trash random
resultset names.

Regardless of how quickly they disappear, or how obscurely they're named,
without an identifier which uniquely identifies the connection to which
the result set belongs, they can be subverted.

Like Microsoft's "Security through Obscurity", this is no security at all.


      ,'/:.          Rob Sanderson ([log in to unmask])
  ,'--/::(@)::.      Special Collections and Archives, extension 3142
,'---/::::::::::.    Twin Cathedrals:  telnet: 7777
____/:::::::::::::.              WWW: