S P Arif Sahari Wibowo wrote:
> On Jul 22, 1:04 pm, Dave Balderstone
> <dave@N_O_T_T_H_I_Sbalderstone.ca> wrote:
>> In article <1185121572.830154.241...@q75g2000hsh.googlegroups .com>, S P
>> Arif Sahari Wibowo <arifs...@gmail.com> wrote:
>>> I tried AppleShare TCP/IP connection from the older Mac, but when
>>> copying, the connection always dropped out after the 1st file.
>> What version of 10.4, and what OS on the 7100?
>
> Sorry I missed this details. The MacOS-X is 10.4.9. The PowerMac 7100
> has one partition with 8.1 and another one running 9.0.4. Both have
> the same issue.
>
> I think I also has MacOS 8.5 CD somewhere in my garage sale stash,
> which I can find and install if it can make a difference.
>
The link that was posted here earlier
http://docs.info.apple.com/article.html?artnum=305420
resonated with me as there have been similar weirdnesses with the latest
LINUX kernel releases.
I can't say that the following is an exact description of the issues,
but its right in general terms.
Essentially the rise in everyones bandwidth has led to usage of some
slightly arcane areas of the TCP/IP specification, mostly around the
area of the window size..the amount of data in packets that will be sent
before the sending machine expects an acknowledgement.
This is supposed to be negotiated by the sending and receiving machine,
but machines in between (typically routers) also can interject and
request smaller sizes.
It seems that many older implementations on older machines and routers
do not work correctly with the modern kernels: The workaround is to
reconfigure the TCP/IP packet sizes to limit their size..whilst this
results in a performance hit on gigabit Ethernet and the like, it
normally has little impact on normal usage.
The symptoms are generically timed out connections during large sized
transfers..
Use of the netstat and ifconfig tools will normally show unusually high
packet loss.
The posted link shows that by adjusting the apple file protocol tunings,
some gains are to be had..this is NOT the same was what I have been
describing, but its very similar.
In practice depending on what you are connecting to, if any data
overruns a buffer, you will get strange effects like those noted..and
experience shows that even a different hub or switch may make things
work better.
> Thanks!
>
> - http://www.arifsaha.com/
>