On Tue, 15 Jun 2004, Carl Brewer wrote:

> Frederick Bruckman wrote:
>> On Tue, 15 Jun 2004, Carl Brewer wrote:
>>> 
>>> Jun 15 11:59:19 fw pppd[165]: write: No buffer space available
>> 
>> I've seen it too. A work-around seems to be to disable compression,
>> or reduce it -- with "deflate 9", I don't get that anymore at all.
>> The problem may be fixed in current and NetBSD 2.0, as they have a
>> newer version of pppd, but I haven't tried them yet for this.
>
> I'll give that a try and see how it goes.  Won't that kill
> some performance though?

Probably not much. I think it means to look for repeating patterns in 
chunks of 512 bytes, rather than chunks of 32 Kb. Most ISP's and POP 
providers weren't even supporting deflate compression until recently, 
which may explain why it wasn't a problem before. I could be way off 
base; check in the log to see if it's doing deflate compression before 
bothering with tweaking it.

>> It's interesting that you're using "passive persist" instead of
>> "demand". Does the other side call you?
>
> No.  This is a config that I inherited and haven't looked
> overly closely at.  Would I be better off with
>
> demand
>
> instead of
>
> passive
> persist

With demand, the daemon will redial if the LCP timeout kicks in, 
or if the connection drops. That would tend to make it more robust. 
However, if the ISP terminal server goes out to lunch while the phone 
line remains OK, it could get expensive (assuming you pay by the call, 
not the minute). For that, you could add a "holdoff" to limit the 
damage.

Frederick