OpenBCM V1.13 (Linux)

Packet Radio Mailbox

DB0FHN

[JN59NK Nuernberg]

 Login: GUEST





  
ZL3AI  > APRDIG   07.02.03 12:01l 283 Lines 9635 Bytes #999 (0) @ WW
BID : 852-ZL3AI
Read: GUEST
Subj: TAPR Digest, Feb 03, 2/6
Path: DB0FHN<DB0RGB<OE5XBL<OE3XSR<OK0PBX<OK0PHL<OK0POV<OK0PKR<OM0PBC<HA5OB<
      HA3PG<HB9OK<IR0NET<VK3TE<ZL2BAU<ZL3VML
Sent: 030205/0845Z @:ZL3VML.#80.NZL.OC #:58833 [Chch-NZ] FBB7.00i $:852-ZL3AI
From: ZL3AI@ZL3VML.#80.NZL.OC
To  : APRDIG@WW

Subject: Re: New spec - working notes
From: Hamish Moffatt <hamish@cloud.net.au>
Date: Tue, 4 Feb 2003 00:23:41 +1100
X-Message-Number: 2

On Sun, Feb 02, 2003 at 11:03:36PM -0000, Miller Scott Contr 30SCS/FTI wrote:
>+/- 128 C is probably enough range, but not enough resolution.  I've got
>temperature sensors with accuracy to better than 1 degree C, and resolution
>to 0.005 C (Intersema 5534).  Differential GPS, with a local reference
>transmitter, can provide centimeter accuracy for position and elevation.
>Yeah, it's not exactly necessary for what we're trying to do here, but for
>simplicity of encoding and decoding I'm trying to stick to an 8-bit
>alignment.

Even with more precision, Kelvin will require more bits to express the
same thing. You could use a range of +/- 64 C (hopefully sufficient!)
and have a fractional bit giving you 0.5 C resolution. 

There are some other arrangements that would make good use of the bits
too. If you could get the required range down to 64 C (eg -20 to +43?)
you could have two fractional bits for 0.25 C resolution. +43 would not
be enough though - we had a 44.5 C degree day here in Melbourne
(Australia) just over 1 week ago.

I think 0.5 C would be more than enough though given that we are not
trying to communicate our lab results using APRS!

I'm enjoying this discussion. Keep at it!

Cheers
Hamish
-- 
Hamish Moffatt VK3SB <hamish@debian.org> <hamish@cloud.net.au>

----------------------------------------------------------------------

Subject: Re: Dictionary of Units of Measurement, now Who said that?
From: Mark Zimmerman <w1bmw@mac.com>
Date: Mon, 3 Feb 2003 08:35:08 -0500
X-Message-Number: 3

>Subject: Re: Dictionary of Units of Measurement - handy reference
>From: Neil Johnson <njohnsn@iowatelecom.net>
>Date: Sun, 2 Feb 2003 08:56:39 -0600
>X-Message-Number: 7

>As I heard someone say, "My car gets 40 rods to the hogshead!".

I don't know how many of the list are Simpsons fans, but that was from 
Homer Simpson's father, Abe.

 /zim - W1BMW - South River, NJ

----------------------------------------------------------------------

Subject: New Dynamic Filter Access
From: "AE5PL Lists" <HamLists@ametx.com>
Date: Mon, 3 Feb 2003 08:36:44 -0600
X-Message-Number: 4

The latest release of javAPRSSrvr now supports an additional APRS
messaging method for clients to activate/modify server filter parameters
on the fly.  In addition to being able to send the filter definition in
an APRS message to the server's callsign, you can also send it to SERVER
This should simplify using the dynamic filter feature via APRS messaging
as you no longer need to know the server's callsign.  This feature does
not replace the messaging to the server callsign method; it is in
addition to.

Servers that currently have the dynamic filter enabled include (all are
currently using port 14580 for this function):

aprsca.net
aprseast.net
aprswest.net
ahubswe.net
second.aprs.net

Other servers may also be supporting dynamic filtering.  Check the
status page of each server for more information on what ports are
available.

73,

Pete Loveall AE5PL
pete@ae5pl.net

----------------------------------------------------------------------

Subject: Re: Future spec
From: "Gregg G. Wonderly" <gregg@skymaster.cytetech.com>
Date: Mon, 03 Feb 2003 08:55:28 -0600
X-Message-Number: 5

>BTW, we have changed the spec to indicate the use of big-endian numbers now,
>to match the Internet standard and the ntohs() and ntohl() functions.

But this makes libraries still not possible, because every compiler and 
platform will consider types differently, including alignment boundries, 
optomized for hardware access (event addresses, quad alignment etc).

The only way that libraries can be written is to create functions for every 
type of value stored that uses byte by byte extraction to get the values out.  
And, don't forget signed vs unsigned bytes and 16 bit bytes...

-----
gregg@cytetech.com  (Cyte Technologies Inc)

----------------------------------------------------------------------

Subject: RE: JAVA-capable Radio
From: "Gregg G. Wonderly" <gregg@skymaster.cytetech.com>
Date: Mon, 03 Feb 2003 09:23:23 -0600
X-Message-Number: 6

>It'd be cool, but a potential nightmare... my brand new Motorola T720 has
>locked up on me several times, requiring a reset.  My Yaesu FT-203 with its
>thumbwheel frequency selector, on the other hand, would probably survive a
>nuclear war.

Most people don't yet understand the differences that Java provides in
system stability and programmer relief of house keeping that plagues larger
embedded systems.  Garbage collection makes sure that casual memory
allocations are returned to the available pool when all references are
dropped.  There are no pointers in Java, so you can't misuse memory
addresses where you have freed memory, but forgot that a pointer somewhere
still references this now free memory.  In java, the GC would not free this
memory because of the outstanding reference.  The most important thing is
exception handling.  You can keep a system alive in the presence of all
kinds of problems by just doing

	public void mainLoop() {
		while(true) {
			try {
				processEvents();
			} catch( Throwable ex ) {
				reportException(ex);
			}
		}
	}

Out of memory errors, and all kinds of problems can crop up in strange 
environments that your software design failed to account for.  Rather than 
being an idiot and believing that you could possibly write 100% correct 
software, you just include the above constructs at the top of all long
running threads and you have made sure that your system will stay running.

Now, there are other software designs that help this too.  Rather than
polling for events, you really should use external stimulous to trigger
work threads to be created and/or run.  That way, if the run into problems
and are terminated by the system, the next event will still be processed.

In typical C based embedded systems, memory management, and infinite loops
due to not using timeouts (keepalive or deadlock detectors etc) are the
primary ways that hard crashes occur.  Java solves the memory problem
trivially. Timing management in Java is also integral.  The base Object of
java includes four methods of interest, wait(), notify(), wait(millis) and
notifyAll().  In a thread that is waiting for another thread to do
something, we can code

	Object lock = ...;

	synchronized(lock) { // Monitor entry
		lock.wait();  // Wait for notification...
	}

and in someother thread, the same lock object would be referenced in the 
context of:

	synchronized( lock ) {  // Monitor entry
		lock.notify();  // Notify waiting thread.
	}

Now, this is the simplistic view.  In reality, the two threads are probably 
not coupled by time, and probably need some external controls.  So, you
could imagine this coded instead as.

	Object txlock = new Object();
	boolean tx = false;

	synchronized(txlock) {
		while( tx ) {
			txlock.wait();
		}
		tx = true;
	}
	... do my TX work

	synchronized(txlock) {
		tx = false;
		txlock.notifyAll();
	}

Thus, we can create a monitor that controls access to the transmitter.
What this affords is that we don't have to have a TX data queue, instead we
implement the queue using the context that is already in place based on 
threads that are doing work which require use of the transmitter.  Now, the 
deal is that we don't want to wait forever for the transmitter.  What if
the transmitting thread dies?  There are two things possible.  We can (and
should always) use try{}finally{}.

	try {
		... do my TX work
	} finally {
		synchronized( txlock ) {
			tx = false;
			txlock.notifyAll();
		}
	}

This will make sure that we always unlock the tx lock no matter what
happens in the try-block.  As well, the hardware control of the tx line
should be released in the finally-block.  Now if there is a way that the
thread with the transmitter might get stuck, we don't want to hold up the
whole system.  So, we can use wait(millis) instead, with a timeout value in
milliseconds that would allow us to know that we are not being woke due to
tx completion.

	Object txlock = new Object();
	boolean tx = false;
	Thread txThread = null;
	long locktime = 5000; // 5 seconds max TX time.
	synchronized( txlock ) {
		while( tx ) {
			txThread = lastTx;
			txlock.wait( locktime );
			// Did we timeout, or does someone else have tx?
			if( tx == true ) {
				if( txThread == lastTx ) {
					// Check for lockup and recover
					tx = checkForTxLockup();
				}
			}
		}
	}

	try {
		txThread = Thread.currentThread();
		... do TX work ...
	} finally {
		synchronized( txlock ) {
			txThread = null; // not current thread now.
			tx = false;
			txlock.notifyAll();
		}
		}
	}

Okay, enough of the introduction to java lesson...

-----
gregg@cytetech.com  (Cyte Technologies Inc)

----------------------------------------------------------------------

Subject: Re: New Dynamic Filter Access
From: "Dick Stanich" <dick@kb7zva.com>
Date: Mon, 3 Feb 2003 9:49:29
X-Message-Number: 7

>Other servers may also be supporting dynamic filtering.  Check the
>status page of each server for more information on what ports are
>available.

Status pages can be viewed in HTML using most browsers. The newer versions 
of javAPRSSrvr allow server sysops to display their port configurations and 
other information.

http://aprsca.net:14501/
http://aprseast.net:14501/
http://aprswest.net:14501/
http://ahubswe.net:14501/
http://second.aprs.net:14501/

Dick, KB7ZVA
APRSWEST.NET

----------------------------------------------------------------------



Read previous mail | Read next mail


 16.11.2026 04:24:21lGo back Go up