Text 906, 178 rader
Skriven 2007-05-06 09:52:14 av Peter Knapper (3:772/1.10)
Kommentar till text 901 av Maurice Kinal (1:261/38.9)
Ärende: sln_sf v0.1
====================
Hi Maurice,
PK> I have no idea what ftpd was being used as I never ran the server my
PK> end.
MK> Then how do you know it can't handle freqs?
Because ftpd is not a Fidonet standard that I EXPECT to be able to do that and
the service prvoider did not provide instructions on HOW to go about doing
that. They DID provide instructios on how to use FTP for Fidonet Mail pickup
and deposit, but that was all AND it was specific to each FTP environment I
used. I DID try sending a REQ once, but that got swallowed so I assumed FREQ
wasnt implemented in anyway..........;-)
MK> How would you know that it isn't better at pure file
MK> transfers then binkd?
Understanding the logic of how FTP works and its limitations in the area of
Fidonet traffic transfers, means its pretty darned clear to me why FTP has
never become a recognised Fidonet standard. I am not saying one specific
implementation cannot be crafted to work in a usable fashion, but I wonder why
one would bother doing this when there are more integrated solutions available
(BinkP) that perform the FIDONET partof the processing in a much easier to
implement solution. I consider it like the chap who put a 307 V8 into a Fiat
Bambina to go to the supermarket, a bit of an overkill...
MK> If you haven't done it there is no way you could know.
That is true, but then I guess that also means I am not a masochist.....;-)
MK> Bottomline is that my statement that binkd is 1/3 as
MK> efficient stands until you prove otherwise which you
MK> can't.
Nope, it stands no further than simply being YOUR statement, as no-one else has
confirmed its performance based on YOUR criteria yet.......;-)
MK> You can post all the logs in the world from
MK> your client and it proves nothing about the server side.
Its pointless looking at only HALF the issue, one has to consider the bigger
picture (IE others withn Fidonet).
PK> I don't, OS/2 has been my mainstay here since V1.3
PK> was available, and its served me very well.
MK> Sounds great. What does that have to do with binkd?
BinkD runs on OS/2 as just one of its platforms...
PK> That I highly doubt, however I certainly don't have any inclination
PK> to replace anything currently in use with Bash.......;-)
MK> Understood. I've never heard of anyone using bash on
MK> OS/2.
Yes, BASH for OS/2 is availabe on Hobbes (bash-2.0-b264.zip), however its a
little old (1997)...
MK> Rexx seems to be the winner there. How much
MK> longer do you think it'll last now that there is little
MK> to nothing going on in OS/2 world?
Obviously you are not up to date on OS/2 events........;-) Yes, IBM has stopped
marketting of OS/2 (but still supported for selected paying environments),
however it is still being developed and marketed as eCS (www.ecomstation.com),
and IBM provided updates to eCS in Feb 2007. You can pick up a copy of the
latest eCS v2.0 Beta (release 4) while you are there...
MK> Right. So in other words that is restricted to that
MK> particular ftpd and not to ftpd in general. It proves
MK> nothing wrt my claim comparing file transfers between
MK> binkd and ftpd, in particular vsftpd and binkd. Right?
So is vsftpd the only FTP server to be considered here? Any idea how many
Fidonet specific implementations of that are being used?
MK> For sure. No doubt about that here. However from my particular
MK> perspective I can see the difference simply because I pick and choose
MK> what will be running on my boxes and I build them from
MK> source optimized to the machine they are meant to run
MK> on.
From that it is clearly obvious to me that what you are looking for are
attributes of your selected environment that most Fidoneters have little
interest in. Fidonet is about moving mail, the by far the largest delays
experienced by Fidoneters is in the transfers bewteen hosts.
In the days of PSTN connectivity most Fidoneters looked strongly towards the
connectivity performance, so Communications protocol performance was important
to them and with the impementation of Fido over IP, the emphasis has altered
only slightly (mainyl due to PSTN costs vs Internet costs), however what does
remain is the Mail processing task, that has not changed.
If the Fidonet S/W environment performs so badly for Fidoneters that they
achieved just 1/3 the throughput that you claim, surely you would expect to
hear something about this? The simple answer is that "it can't be happening to
most Fidoneters that way"...
EG: All my Fidonet processing is done on the SMALLEST machine I have here. Its
the smallest machine that I have because when I purchased the replacment for a
failed 486SX-20 motherboard in 2001 the LEAST powerful machine I could buy new
was a Celeron 700 with 64MB RAM! That machine has run Warp 4, Maximus, Squish,
BinkD, Apache, SyslogD, a Time server, heaps of Rexx scripts for mail handling
for over a decade (IE it did this on the 486SX-20 with 16MB RAM!). When it was
on the 486SX-20 it ran 4 modem lines using a Digiboard, down to just 1 PSTN
line now. All this has run fine for me, however I can assure you that
operational costs are a definite factor to me.
As for performance, there has been near ZERO change in BinkD and FTP
performance for the BBS operations between the 486DX2-66 and the Celeron. The
only logical conclusion that I can come to is that when I run FTP and BinkD on
the same machine to do similar (but not exactly the same because they can't DO
exactly the same) jobs, I see EXACTLY the performances I expect to see, and
optimise my handling of the traffic to take advantage of that.
Even you must be able to see that if FTP was so wonderful (3 times faster than
BinkP), then Fidoneters would flock to it in droves. The reality is that they
don't, so your perceived "advantage" of using FTP must be pretty small to most
Fidoneters, considering the efforts put into optimsing things like ZEDZAP and
other PSTN protocols.
MK> Exactly where is the hobby
MK> in that mentality? This I would really appreciate knowing.
The Hobby mentality is focused on costs and performance, but by perfromance I
mean "how does it perform its Fidonet functions", and the answer so far is
clearly that the existing Fidonet protocols (IE BinkP) work fine.
PK> speed of transfer is not an issue to me, as much of that is totally
PK> outside my control.
MK> Exactly. It is here to a greater extent and I am
MK> prepared to do something about it or at least how it
MK> pertains to my part.
I dont see how any of us can do this, because its all "caused" by the internet
and that is well outside of our "control".
MK> Beyond what is in my control has
MK> nothing to do with me. However both binkd and vsftpd
MK> are and I reiterate my original claim about efficiency.
I dont doubt that you consider there to be a significant performance benefit,
all I am doing is asking is "why have no others seen this same 3 x speed
improvement" that you have, because I certainly don't see it...
PK> my Rexx FTP code does pretty much EXACTLY what I want it to do, why
PK> should I even consider changing it, if I am not changing platforms?
MK> I wouldn't if I were you. But I am not you and I don't
MK> have OS/2. Isn't it abandonware or close to it?
See above......;-)
MK> We'll talk in a few years from now and compare notes.
If Fidonet continues for that long and we are still around.....;-)
Cheers.............pk.
--- Maximus/2 3.01
* Origin: Another Good Point About OS/2 (3:772/1.10)
|