[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [help-GIFT] viper/gift log files

From: David Squire
Subject: Re: [help-GIFT] viper/gift log files
Date: Mon, 05 Mar 2001 11:31:35 +1100

Wolfgang Mueller wrote:

> >>>>> "HM" == Henning Mueller <address@hidden> writes:
> HM> When the client opens a session, the user name is transmitted with a
> HM> dummy session-number. Then the server transmits the real session number,
> HM> but no user-name. Is it guaranteed that these two messages are always
> HM> right after each other in the log file? Otherwise it is not guaranteed,
> HM> that the connection between session-id and user-name can be made. This
> HM> did not cause any problem for me, but may be it could.
> This is a good question.
> In fact, it's yes and no:
> In theory, it can happen, in practise it will only happen at very high
> load, as once the accessors are initialised, opening a session
> currently is VERY fast. Opening a session means creating a CQuery, and
> attaching a CAccessor to it. This can happen much more quickly than a
> blink of an eye, so there would have to be another query which is sent
> exactly in this moment. Since gift is multithreaded as of 0.1.5pre,
> this *could* happen, but is very improbable.

And thus, I think, that it is worth ensuring that it can't happen.

> HM> Wolfgang, do you think that it would make sense to send the user name
> HM> back in the message with the real session-id? Like this we would have a
> HM> direct connection between the two.
> For reasons of redundancy I would be against *sending* that. What I
> would suggest would be *writing* this information into the
> logfile. The same is the case for timing and server information. I
> would suggest to put in front of each message a comment which further
> describes the properties of the message. Or to add to the mrml element
> that has been sent an XML element containing connection information,
> after sending, but before writing all this to the log file.
> things would look afterwards like that in the log file:
> ...
> <mrml ...>
>       <cui-connection-information name="henning"
>                                   ip=""
>                                   date="Fri Mar  2 10:56:15 CET 2001"/>
>       ...
> </mrml>
> ...
> This would make it as easy as possible to get at the information when
> parsing the log file.
> What think?

Putting this stuff in the log file is fine, but I can't see why you wouldn't
send it. One of the nice things about the MRML logs (as opposed to my earlier
ad hoc ones) is that they are a true record of what actually passed between
client and server. I would not like to see that principle changed.

Question: what happens if there are multiple sessions started from the same IP?
How is this resolved? What if there are multiple sessions from the same IP with
the same user name?



Dr. David McG. Squire
Computer Science and Software Engineering, Monash University, Australia
Do/Don't want HTML mail? Let me know.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]