D
dave hillstrom
Guest
On Wed, 13 Aug 2008 02:26:23 -0400, mimus <tinmimus99@hotmail.com>
wrote:
>On Tue, 12 Aug 2008 23:30:47 -0400, mixed nuts wrote:
>
>> mimus wrote:
>>
>>> On Tue, 12 Aug 2008 20:28:56 -0400, dave hillstrom wrote:
>>>
>>>> On Tue, 12 Aug 2008 06:20:03 -0400, mixed nuts
>>>> <melopsitticus@undulatus.budgie> wrote:
>>>>
>>>>> dave hillstrom wrote:
>>>>>
>>>>>> On Mon, 11 Aug 2008 18:52:49 -0400, mimus <tinmimus99@hotmail.com>
>>>>>> wrote:
>>>>>>
>>>>>>> On Mon, 11 Aug 2008 08:56:00 -0400, dave hillstrom wrote:
>>>>>>>
>>>>>>>> On Sun, 10 Aug 2008 19:50:49 -0400, mimus <tinmimus99@hotmail.com>
>>>>>>>> wrote:
>>>>>>>>
>>>>>>>>> On Sun, 10 Aug 2008 17:11:31 -0400, mixed nuts wrote:
>>>>>>>>>
>>>>>>>>>> dave hillstrom wrote:
>>>>>>>>>>
>>>>>>>>>>> On Sun, 10 Aug 2008 12:26:56 -0400, mimus
>>>>>>>>>>> <tinmimus99@hotmail.com> wrote:
>>>>>>>>>>>
>>>>>>>>>>>> On Sun, 10 Aug 2008 01:55:56 -0400, dave hillstrom wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> On Sun, 10 Aug 2008 01:39:15 -0400, dave hillstrom
>>>>>>>>>>>>> <DaVe@MeOw.OrG> wrote:
>>>>>>>>>>>>>
>>>>>>>>>>>>>> combinatorics has been running through my noggin for the
>>>>>>>>>>>>>> last few hours on a topological problem that i really want
>>>>>>>>>>>>>> to forget and relax from. but it intrudes. feh, i say.
>>>>>>>>>>>>>> feh.
>>>>>>>>>>>>>
>>>>>>>>>>>>> ~n~ ~matrixed~ combinatorics, at that!!!!1!!
>>>>>>>>>>>>
>>>>>>>>>>>> Just try to keep one limb on an orthogonal somewhere.
>>>>>>>>>>>
>>>>>>>>>>> ok, ok. heres a real network problem im up against, as well.
>>>>>>>>>>> ive got a solution (pushbutton on slaves to announce
>>>>>>>>>>> themselves), but its not as elegant as i want.
>>>>>>>>>>>
>>>>>>>>>>> say you have a bunch of units, each with a unique 4 byte
>>>>>>>>>>> address, programmed in at production time (much like network
>>>>>>>>>>> cards MAC addresses). say you can define one as master and you
>>>>>>>>>>> can connect up to 31 slaves to it. what is the fastest way on
>>>>>>>>>>> powerup for the master to discover whos on its network? keep
>>>>>>>>>>> in mind that the network physical layer is half-duplex rs485
>>>>>>>>>>> running at 57.6Kbaud.
>>>>>>>>>>>
>>>>>>>>>>> got any ideas?
>>>>>>>>>>
>>>>>>>>>> Wait 'til the design review meeting - you'll get at least 10
>>>>>>>>>> suggestions that will be much faster than you imagined possible.
>>>>>>>>>
>>>>>>>>> Let the slaves check in to the master when they need to.
>>>>>>>>
>>>>>>>> according to what schedule? their ALL going to want to check in
>>>>>>>> at startup if i use an auto discovery scheme. and with that
>>>>>>>> address space at 57kbaud it would take forever if they use a
>>>>>>>> linear time delay derived grom their address.
>>>>>>>
>>>>>>> What about letting 'em keep PINGing the master with their addresses
>>>>>>> until it finally answers each? you can lock the master I/O latch
>>>>>>> against input, one presumes.
>>>>>>>
>>>>>>> Can you hook 'em in a ring and pass a sign-up sheet around
>>>>>>> ("<MASTER ADDRESS:>XXXX<APPEND YOUR ADDRESS TO THIS LIST:>")?
>>>>>>
>>>>>> no rings. strictly chaos.
>>>>>>
>>>>>> pinging the master at some fixed interval would do no good, as
>>>>>> others will do the same.
>>>>>
>>>>> Yore stuck walking down a binary tree then. Or up, depending on yore
>>>>> prespective and how you decide to order the search. With amazing
>>>>> luck you could complete the mapping in mn cycles (where m is the
>>>>> total number of addresses possible and n is the number of slaves
>>>>> actually connected) but most of the time you'll have to resolve
>>>>> collisions and it will take longer, but not longer than 2^m-1.
>>>>> Assuming 5 k bytes/sec and 4 bytes/yesno transaction and a 64k
>>>>> address space, you'll be mapped in less time than it'll take to boot
>>>>> windows - with or without tomateos.
>>>>
>>>> too bad its a 32 bit address space. though that might not change the
>>>> time too too much....
>>>
>>> Don't you _know_ the damned things' addresses when you hook 'em up,
>>> allowing for some nice EPROM-array or even text-file-based polling on
>>> power-up?
>>>
>>> Kludgy, but not as kludgy as stepping through four million possible
>>> addresses-- on a serial line, at that <shudder>-- or wotever . . . .
>>
>> Yuo mis-peeled billion. Or has Ohio joined the British Commonwealth?
>> hmmmmm?
>
>I don't know what I mean.
>
>There's always jumper-set addresses, if you really want to get, like,
>paleo-kludgy . . . .
>
>(Reminds me of the paper-clip I used to use to reboot my C64s with.)
too much to ask for from the user, unfortunately. i mean, a lot of
these folks are gonna be dirt stupid.
--
dave hillstrom mhm15x4 zrbj
"i believe that the word "****head" has become so wide spread and
nearly meaningless as to qualify as a metavariable, similar to "foo"
and "bar". and that it should uphold the responsibilities and enjoy
the privileges of the new office. here here!!"
-dave hillstrom
wrote:
>On Tue, 12 Aug 2008 23:30:47 -0400, mixed nuts wrote:
>
>> mimus wrote:
>>
>>> On Tue, 12 Aug 2008 20:28:56 -0400, dave hillstrom wrote:
>>>
>>>> On Tue, 12 Aug 2008 06:20:03 -0400, mixed nuts
>>>> <melopsitticus@undulatus.budgie> wrote:
>>>>
>>>>> dave hillstrom wrote:
>>>>>
>>>>>> On Mon, 11 Aug 2008 18:52:49 -0400, mimus <tinmimus99@hotmail.com>
>>>>>> wrote:
>>>>>>
>>>>>>> On Mon, 11 Aug 2008 08:56:00 -0400, dave hillstrom wrote:
>>>>>>>
>>>>>>>> On Sun, 10 Aug 2008 19:50:49 -0400, mimus <tinmimus99@hotmail.com>
>>>>>>>> wrote:
>>>>>>>>
>>>>>>>>> On Sun, 10 Aug 2008 17:11:31 -0400, mixed nuts wrote:
>>>>>>>>>
>>>>>>>>>> dave hillstrom wrote:
>>>>>>>>>>
>>>>>>>>>>> On Sun, 10 Aug 2008 12:26:56 -0400, mimus
>>>>>>>>>>> <tinmimus99@hotmail.com> wrote:
>>>>>>>>>>>
>>>>>>>>>>>> On Sun, 10 Aug 2008 01:55:56 -0400, dave hillstrom wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> On Sun, 10 Aug 2008 01:39:15 -0400, dave hillstrom
>>>>>>>>>>>>> <DaVe@MeOw.OrG> wrote:
>>>>>>>>>>>>>
>>>>>>>>>>>>>> combinatorics has been running through my noggin for the
>>>>>>>>>>>>>> last few hours on a topological problem that i really want
>>>>>>>>>>>>>> to forget and relax from. but it intrudes. feh, i say.
>>>>>>>>>>>>>> feh.
>>>>>>>>>>>>>
>>>>>>>>>>>>> ~n~ ~matrixed~ combinatorics, at that!!!!1!!
>>>>>>>>>>>>
>>>>>>>>>>>> Just try to keep one limb on an orthogonal somewhere.
>>>>>>>>>>>
>>>>>>>>>>> ok, ok. heres a real network problem im up against, as well.
>>>>>>>>>>> ive got a solution (pushbutton on slaves to announce
>>>>>>>>>>> themselves), but its not as elegant as i want.
>>>>>>>>>>>
>>>>>>>>>>> say you have a bunch of units, each with a unique 4 byte
>>>>>>>>>>> address, programmed in at production time (much like network
>>>>>>>>>>> cards MAC addresses). say you can define one as master and you
>>>>>>>>>>> can connect up to 31 slaves to it. what is the fastest way on
>>>>>>>>>>> powerup for the master to discover whos on its network? keep
>>>>>>>>>>> in mind that the network physical layer is half-duplex rs485
>>>>>>>>>>> running at 57.6Kbaud.
>>>>>>>>>>>
>>>>>>>>>>> got any ideas?
>>>>>>>>>>
>>>>>>>>>> Wait 'til the design review meeting - you'll get at least 10
>>>>>>>>>> suggestions that will be much faster than you imagined possible.
>>>>>>>>>
>>>>>>>>> Let the slaves check in to the master when they need to.
>>>>>>>>
>>>>>>>> according to what schedule? their ALL going to want to check in
>>>>>>>> at startup if i use an auto discovery scheme. and with that
>>>>>>>> address space at 57kbaud it would take forever if they use a
>>>>>>>> linear time delay derived grom their address.
>>>>>>>
>>>>>>> What about letting 'em keep PINGing the master with their addresses
>>>>>>> until it finally answers each? you can lock the master I/O latch
>>>>>>> against input, one presumes.
>>>>>>>
>>>>>>> Can you hook 'em in a ring and pass a sign-up sheet around
>>>>>>> ("<MASTER ADDRESS:>XXXX<APPEND YOUR ADDRESS TO THIS LIST:>")?
>>>>>>
>>>>>> no rings. strictly chaos.
>>>>>>
>>>>>> pinging the master at some fixed interval would do no good, as
>>>>>> others will do the same.
>>>>>
>>>>> Yore stuck walking down a binary tree then. Or up, depending on yore
>>>>> prespective and how you decide to order the search. With amazing
>>>>> luck you could complete the mapping in mn cycles (where m is the
>>>>> total number of addresses possible and n is the number of slaves
>>>>> actually connected) but most of the time you'll have to resolve
>>>>> collisions and it will take longer, but not longer than 2^m-1.
>>>>> Assuming 5 k bytes/sec and 4 bytes/yesno transaction and a 64k
>>>>> address space, you'll be mapped in less time than it'll take to boot
>>>>> windows - with or without tomateos.
>>>>
>>>> too bad its a 32 bit address space. though that might not change the
>>>> time too too much....
>>>
>>> Don't you _know_ the damned things' addresses when you hook 'em up,
>>> allowing for some nice EPROM-array or even text-file-based polling on
>>> power-up?
>>>
>>> Kludgy, but not as kludgy as stepping through four million possible
>>> addresses-- on a serial line, at that <shudder>-- or wotever . . . .
>>
>> Yuo mis-peeled billion. Or has Ohio joined the British Commonwealth?
>> hmmmmm?
>
>I don't know what I mean.
>
>There's always jumper-set addresses, if you really want to get, like,
>paleo-kludgy . . . .
>
>(Reminds me of the paper-clip I used to use to reboot my C64s with.)
too much to ask for from the user, unfortunately. i mean, a lot of
these folks are gonna be dirt stupid.
--
dave hillstrom mhm15x4 zrbj
"i believe that the word "****head" has become so wide spread and
nearly meaningless as to qualify as a metavariable, similar to "foo"
and "bar". and that it should uphold the responsibilities and enjoy
the privileges of the new office. here here!!"
-dave hillstrom