Ethereal-dev: Re: [Ethereal-dev] Menu discussion (was 0.10.2 release)
Note: This archive is from the project's previous web site, ethereal.com. This list is no longer active.
From: "Olivier Biot" <ethereal@xxxxxxxxxx>
Date: Sun, 22 Feb 2004 23:04:06 +0100
From: Ulf Lamping | Olivier Biot wrote: | | >From: Ulf Lamping | > | >| Richard Sharpe wrote: | >| | >| >On Sun, 22 Feb 2004, Ulf Lamping wrote: | >| > | >| >>Hmmm, I just checked in some major rework of the menu structure of | >| >>Analyze/Statistics: | >| >> | >| >>Sorted by ISO-layer then by alphabetical order. | > | >Here I am again, nitpicking on ISO-OSI :) In some environments it is | >not really obvious to talk about a given OSI layer. This is | >particularly true in areas where clsssical telecom and the Internet | >are joined. I tend to define (parts of) the OSI stack with a starting | >reference point. To state it in a simple manner: user A's application | >layer is maybe user B's transport layer. For this reason, I feel | >unhappy about using absolute OSI reference layer mappings to | >protocols. | | I don't think that's true (at least on the protocols we currently have | taps for), | e.g. the IP protocol is a network protocol, regardless what's above and | what's below it and what you are doing with it in your specific task. Umm... maybe I am 100% wrong but if I remember correctly from some discussions with 3G (aka UMTS) people, the IP layer is the *application* layer when we're dealing with say the IMS, while for me, looking at the IP on top of this IMS low-level stuff, IP is indeed the *network* protocol. Of course, I agree that if you *only* look at IP, then the IP protocol's use is as a network protocol. | It might be true that you have multiple network protocols | in your stack, but that doesn't change the facts. I get your point. You would propose to add *all* network protocols in the *OSI network layer* menu item, and do the same for all OSI layers. That sounds feasible today as we only have a limited set of protocols for which we generate statistics. Doing so would already offer some more logic in the menu than what is available today, however it does not allow a user to see the OSI stack in which a given protocol fits. We then still need to deal with situations where stats are being generated for more than one protocol at the same time. And that's a question I have right now: do we want to sort the items in an OSI stack manner, or do we want to do it in a per-OSI-stack family (telecom signalling, IP, etc)? We can probably only answer this question by trying out one way, and then evaluate the situation with end-user feedback. | There might be other protocols, where it's not that sure. | And keep in mind: *if* you are doing this and trying to analyze things, | you will usually *know* that you're doing | "unusual things" with it, and what the usual "use case" would be (so | where you have to look at). Here I lost the track, sorry. | >| The previous structure was the other way round, you had to have a | >| function in mind, | >| and then look after the protocols implemented this function. | > | >I think both approaches are valid. However an end-user might expect | >some functionality which is not present for the desired protocol | >because it was not implemented (yet). | > | >I think we should go for a mix of both approaches. Maybe we could have | >the taps register the protocol name of every protocol they act upon so | >we could use context-specific menus when we select a given packet or a | >given protocol field. Maybe we need to define *some* protocol | >hierarchy (like a (partial) OSI layered approach). Maybe we want to | >separate the protocol matter in some sort of "profiles" like telecom, | >Internet etc. Maybe we want to do a mix of all of this :) | > | Sorry, but I'm pretty sure you are really doing overkill on this topic | right now!!! Honestly, I don't see a problem in listing options as a hint for the future, while you are working *now* on a fix which can be implemented in a fair amount of time. Sometimes this approach helps in finding a (sometimes obvious) solution. | Context specific menus! Do you really have the time to implement this? I guess not. Besides, I still need to read much GTK code before I attempt at doing one of the maybe esoteric changes I proposed, and which were mainly meant as input for a discussion for some long-term solution. | I wanted to find a solution which could be implemented in a reasonable | amount of time. This is 100% clear to me and makes sense to the same extent :) However, what I wanted to *suggest*, is (a) that we may want to sort the menu items in application categories, such as "IP", "telephony", "VoIP", "ATM", "Frame Relay" etc in order to define the menu trees in a more intuitive manner. This has also been suggested by Michael Tuexen. Of course, intuition sometimes is something of a personal matter :) Additionally, (b) we might want to look at one or more long-term solutions which (I agree) might or might not be relevant anymore after we've seen the short-term solution. | There are many other things in the GUI that are in definite need for | some rework too ... Agreed. But let's discuss only the menu stuff for the stats for now. I am happy to see Ethereal evolving like it is today! Regards, Olivier
- References:
- Re: [Ethereal-dev] 0.10.2 release
- From: Richard Sharpe
- Re: [Ethereal-dev] 0.10.2 release
- From: Ulf Lamping
- [Ethereal-dev] Menu discussion (was 0.10.2 release)
- From: Olivier Biot
- Re: [Ethereal-dev] Menu discussion (was 0.10.2 release)
- From: Ulf Lamping
- Re: [Ethereal-dev] 0.10.2 release
- Prev by Date: Re: [Ethereal-dev] bug submission ...
- Next by Date: [Ethereal-dev] Re: [Ethereal-cvs] cvs commit: ethereal README.win32
- Previous by thread: Re: [Ethereal-dev] Menu discussion (was 0.10.2 release)
- Next by thread: Re: [Ethereal-dev] 0.10.2 release
- Index(es):