[9fans] 9p2000 query

[9fans] 9p2000 query

[I RATTAN]

Does 9P preservs message boundaries?
(Single write at this end, can be retrieved with single read
on the other end).

-ishwar

Re: [9fans] 9p2000 query

[Russ Cox]

yes.

On 5/13/05, I RATTAN <rattan@cps.cmich.edu> wrote:
> Does 9P preservs message boundaries?
> (Single write at this end, can be retrieved with single read
> on the other end).
> 
> -ishwar
> 
>

Re: [9fans] 9p2000 query

[Russ Cox]

or no, depending on the file server.
it's not really a well-formed question now that i think about it.

devpipe preserves message boundaries.
il does too.
tcp does not.

On 5/13/05, Russ Cox <russcox@gmail.com> wrote:
> yes.
> 
> On 5/13/05, I RATTAN <rattan@cps.cmich.edu> wrote:
> > Does 9P preservs message boundaries?
> > (Single write at this end, can be retrieved with single read
> > on the other end).
> >
> > -ishwar
> >
> >
>

Re: [9fans] 9p2000 query

[andrey mirtchovski]

> or no, depending on the file server.
> it's not really a well-formed question now that i think about it.
> 
> devpipe preserves message boundaries.
> il does too.
> tcp does not.
> 

"Plan 9 from Bell Labs" discusses message boundaries briefly in the
context of 9p:

"The 9P protocol must run above a reliable transport protocol with
delimited messages.  9P has no mechanism to recover from transmission
errors and the system assumes that each read from a communication
channel will return a single 9P message; it does not parse the data
stream to discover message boundaries.  Pipes and some network
protocols already have these properties but the standard IP protocols
do not.  TCP does not delimit messages, while UDP [RFC768] does not
provide reliable in-order delivery."

those may not be the same message boundaries Ishwar was looking for
though.

Re: [9fans] 9p2000 query

[Brantley Coile]

>> or no, depending on the file server.
>> it's not really a well-formed question now that i think about it.
>> 
>> devpipe preserves message boundaries.
>> il does too.
>> tcp does not.
>> 
> 
> "Plan 9 from Bell Labs" discusses message boundaries briefly in the
> context of 9p:
> 
> "The 9P protocol must run above a reliable transport protocol with
> delimited messages.  9P has no mechanism to recover from transmission
> errors and the system assumes that each read from a communication
> channel will return a single 9P message; it does not parse the data
> stream to discover message boundaries.  Pipes and some network
> protocols already have these properties but the standard IP protocols
> do not.  TCP does not delimit messages, while UDP [RFC768] does not
> provide reliable in-order delivery."
> 
> those may not be the same message boundaries Ishwar was looking for
> though.

Do you mean 9p requires message boundaries to function, or is it part
of the protocol?  9p2000 doesn't need it to work, like the previous
version of 9p did.  It's not part of the protocol, since it would be
the transport's job.  Each 9p2000 message has a length field for the
message.  Old 9p didn't and you had to use il or some extra cruft to
get message boundaries over tcp.

Re: [9fans] 9p2000 query

[I RATTAN]

Here is rephrasing:

server process om machine X is servicing tokens to clients (run on
other machines):

client on A sends a request (message), wsits for reply to come back
can server process pickup the request with single read operation
(assume that a connection has been set up between client and
server processes)?

scenarios possible here:
- one can use tcp from tcp/ip (it is there under Plan 9)
- one can use udp from tcp/ip (is it available under Plan 9)
- one can use 9P without worrying about the underlying protocol(s)
  as it will provide guaranteed service??

-ishwar

On Fri, 13 May 2005, andrey mirtchovski wrote:

> > or no, depending on the file server.
> > it's not really a well-formed question now that i think about it.
> >
> > devpipe preserves message boundaries.
> > il does too.
> > tcp does not.
> >
>
> "Plan 9 from Bell Labs" discusses message boundaries briefly in the
> context of 9p:
>
> "The 9P protocol must run above a reliable transport protocol with
> delimited messages.  9P has no mechanism to recover from transmission
> errors and the system assumes that each read from a communication
> channel will return a single 9P message; it does not parse the data
> stream to discover message boundaries.  Pipes and some network
> protocols already have these properties but the standard IP protocols
> do not.  TCP does not delimit messages, while UDP [RFC768] does not
> provide reliable in-order delivery."
>
> those may not be the same message boundaries Ishwar was looking for
> though.
>

Re: [9fans] 9p2000 query

[Ronald G. Minnich]

On Fri, 13 May 2005, I RATTAN wrote:

> client on A sends a request (message), wsits for reply to come back
> can server process pickup the request with single read operation

that is totally dependent on transport, right? for tcp, you have to do a 
little work. That's why the byte count at the front of the 9p2000 messages 
is so nice.

ron

Re: [9fans] 9p2000 query

["Nils O. Selåsdal"]

Russ Cox wrote:
> or no, depending on the file server.
> it's not really a well-formed question now that i think about it.

I was under the impression the iounit also comes into play,
and that this was independant of the transport used (when you're
at the application level, issuing read(2) and write(2)s )?

"The iounit field returned by open and create may be zero. If it is not, it is 
the maximum number of bytes that are guaranteed to be read from or written to 
the file without breaking the I/O transfer into multiple 9P messages;"
"