Re: Proposal: make init optional

Bob

we would need to think a little carefully about some of this

buffer sizes - what happens if the server sends more not less than the
client can handle

services - we sould need some new error mechanisms (perhaps just
diagnostics) to indicate the service is not supported

protocol version - again there are some things you can do in v3 but not in
v2 do the apdu's are laid out a bit differently - but gain diagnostics
could handle that

authenitcation needs to be dealt with

but this could be an interesting discussion - but for the purposes you
want to do away with init - doesnt encapsulation handle a lot of what you
are trying to do?

mark


On Mon, 8 Jan 2001, Robert Waldstein wrote:

> 
> 
> > how would you propose that some of the negotiation about buffer sizes and
> > services supported (plus protocol version) be handled if there is no init
> > - Im not necessarily opposed to this - just interested in how you see
> > those types of things being handled
> 
>  - buffer sizes:  do alot of this negotiation in HTTP?  I just wouldn't
>    bother, you would get what you asked for or as much as I was willing
>    to give you.
> 
>  - services: this actually always bothered me.  
>       What does a server do now with a client that ignores
>       the option bits? Oh yes - if good OSI targets they shut down cause
>       it is a "bad" client.
> 
>  - protocol version: actually this is interesting.  But so far I don't think
>      we use it? That is, PDUs aren't laid out differently in the different
>      versions - they just have new features added.  If we had laid out the
>      PDUs differently this would be a serious problem; but the feature issue
>      I answer like the above; that is, what do the servers do now when
>      something is requested/used that they don't know or support.
> 
>   - authentication (you didn't mention it but someone will):  I never liked
>     this anyway.  I would think most servers have something they wish could
>     be free (e.g. advertising their service, test DBs, etc). My server
>     certainly has DBs with different level of needed authentication.  So
>     yes, if you need authentication do access control.
> 	THis I think people will object to - could also solve this be having
>     and agreed upon otherInfo item for access control.
> 
>   bob waldstein wald@lucent.com
>

Received on Monday, 8 January 2001 10:26:12 UTC