i hate computers so much

[solarbird]

it's not the RAM (passes memtest86), it's not the DRIVER (replaced with newer version), it's apparently not the CARD (replaced) - what the FUCK is making our transfers hang?

What's extra infuriating is the way that it's happening. Transfers go at high speed and then thunk to zero and never recover. Bit 0x20 in the checksum starts being inverted for the packet that is trying to be sent, and it never fixes no matter how many times its retried.

It may and may not be dependent upon data. It seems dependent upon host. (Most people can download this one test file one; one person can't, ever, and sees the abovementioned error. However, that person sees it from multiple hosts.)

It's infuriating, is what it is.

Comments

[firni.livejournal.com]

Preston says: "that size is suspiciously close to the maximum of a 32 bit signed integer. It could be that their application is written in such a way that they're using the signed int to calculate the size and it's a doing a rollover to a negative number. Is there another application she can try that would mimick the same thing?"

ALL THIS IS GREEK TO ME, YOU NERDS

[firni.livejournal.com]

Let me guess:

NO HIRE

[solarbird]

This is at the TCP/IP stack level.

[firni.livejournal.com]

I told him that and he grunted because now my dad has his clutches into Preston for evil purposes of getting downloaded scenarios to work in MS Train Simulator.

NO I'M NOT KIDDING

[shikyrie.livejournal.com]

Doh... I got an A in that class, but damned if I know what it could be... Computers are STILL greek to me... that's why I didn't bother fixing the problem with my old computer and opted to buy a new one...

[ssha.livejournal.com]

Oh, BTW... I couldn't get the FTP server to accept a connection last night. Ended up just using one of my backup webspace accounts.

[solarbird]

We continue not to support ftp. Use sftp or scp instead.

Help here

[backrubbear.livejournal.com]

I think I missed something in the thread.

You've determined that your host is failing to xmit the packet with a correct checksum?

[solarbird]

We've determined that when it gets there, it has an incorrect checksum.

This same transformation is happening inbound as well, we're pretty sure. (The symptoms are identical.)

[backrubbear.livejournal.com]

This is very intriguing.

I encountered a similar problem thorugh a very well known and very short network path in my previous job, but never managed to actually troubleshoot it.

If you were to do a packet capture and replay, will this packet always get there corrupted? I would be very interested.

Also, which checksum is getting hammered? IP or TCP?

[wrog]

I'm reminded of a grad-school friend of mine's account of the Fortran-hating gateway.

[mistwolf.livejournal.com]

Just out of curiosity, have you talked to your ISP about this? Could it be something on THEIR end? Can you try connceting another host entirely to your connection for a bit, and see if it has the same problem? That could isolate it from the actual box in question.