Split FMP file, Value List Best Option?

[TheLagosChap]

Hi,

I'm using a split filemaker application:

- GUI (layouts, reports etc) stored in one FMP file running on a client computer.

- Data stored in another FMP file running on a FM Server.

I've got 2 ways to define a Value List (to show the value of a field in a table)

1. On the client: "Using Values from Field" (pointing thus to a field in a table stored on the server FMP file)

or

2. On the server: define a value list #1 "Using Values from Field" AND on the client: define a value list #2 pointing to that one on the server ("Use value list from another file").

I'm wondering what is the best option in terms of display speed.

It all depends on when/how the value list is updated from the table.

Anyone has clue or experience on any performance difference using one of the other option?

Note: this performance question is also applicable to value lists made of "Custom Values" in case the list of values is very long.

Thanks.

[Søren Dyhr]

I don't know what the answer is to this, if it is measurable at all ...what really matters is the length of the lists - which sometimes better can be served via a portal with the choises, or rather portals each having thier own relationship pending in tabbed views - it's always the realestate that changes that dictates the speed. There are e.g. differences in renderingtime for portals with and without scrollbars attached ...the smaller distinctive logical chunks you can usher in and out, the faster the application responds.

However can't I help finding the way you utilize the separation model is a little penny wise pound foolish. The separation model is mainly a meant to develop within some kind rollback if newly developed module shows peculiar habits not quite expected, requirering very little downtime to switch back.

I have seen your particular use of the separationmodel before, because the developers can't constrain themselves to vectorized graphics only, and urges to plaster the interface with all kind of photoshop'ish nifties, to distract the user from the core of why to choose a particular tool over another ...which is productivity!

If your solution is too slow rendering isn't it my huble opinion the throughput of sets of numeric or textual data, but much more likely the eyecandy thats causing the drawbacks.

Some kind of guidelines can be deducted from this:

http://www.aptworks.com/downloads/presentations/PacketWatchingPres_v2.PDF

and somewhat builing on this, for the newer versions:

http://network.datatude.net/viewtopic.php?t=102

--sd

[TheLagosChap]

Thank you for the links and feedback provided.

Yes of course, lenght of value lists is obviously important.

But sometimes difficult to control or restrict.

Like one I'm using is a list of employees; if the company has 10, it's easy, if the company has 2000, the corresponding drop down list will take much longer to load/display.

The reason why I'm using the separation model is because I'm working on very slow bandwidth network. In Africa in general and Nigeria in particular, connections are VERY expensive (like: 2,000 US$ per month for a 128Kbps line) and you have traffic for emails in and out, users browsing internet etc on top of filemaker usage.

-(

The theories discussing the performance difference when using the separation model or not are very interesting but if you actually use FMP over a low bandwidth network, you can experience it and probably say even more about it.

My applications do NOT use any graphic at all on the layouts and my experience shows that the user won't even want to use the application if it entirely resides on the server because it's much too slow to navigate.

When I use the separation model, the application can actually be used.

That's a reason good enough for me to use the separation model (besides of course the benefit of being able to develop a new version without disturbing the users).

:-)

So this is why I'm trying to find out how FM works internally (for instance, what is loaded into which memory and at what moment, when and how it is transfered between server and client etc) to see if I can implement the best method for my value lists.

Cheers,

LChap.

[Søren Dyhr]

In Africa in general and Nigeria in particular, connections are VERY expensive (like: 2,000 US$ per month for a 128Kbps line) and you have traffic for emails in and out, users browsing internet etc on top of filemaker usage.

Ah we're talking wan'ing, for this particular situation should either Citrix'ing or Syncdek'ing give ROI swiftly!

--sd

[Fenton]

One thing you can do is to design the interface for maximum speed (as you know). In the specific case of value lists I would say do not use value lists for fields with 2000 entries. Use a dedicated filtered portal, probably on a dedicated layout; a new window sized and positioned appropriately works well (though it triggers restored windows on Windows). Put a button, or just click on the row itself to make a choice.

Use either a Custom Function to "explode" the names (for the stored target field), or manually create it (Left (First, 1) & ¶ & Left (First, 2), etc.. Use either a global field, or a user-record's field in a users-table as the originating field (may be faster than a global in these extreme conditions; don't really know). Clear the left side after a choice and close the little window (check that it's not the only window).

This way they see nothing until they make a choice. Use this general idea as much as possible instead of long value lists.

[Søren Dyhr]

table as the originating field (may be faster than a global in these extreme conditions; don't really know)

Yes this is what I've learned somewhere, use globals sparsingly since they're local to each user and causes a lot of chatter in the connection if they're used in a calculation.

Another thing that occured to me, was to utilize this:

http://www.win4lin.com/content/view/195/206/

http://www.sun.com/software/sdis/

To make filemaker GOffice'ish!

--sd