Wrappers to do full sendNetworkData and ReceiveNetworkData

[idle]

This should do it on linux. Need to get it from libc

Code: Select all

ImportC "-lc"
__errno_location() 
EndImport   
error = PeekI(__errno_location()) 

[Quin]

Hi Idle,
You're correct, this seems to work, for the asm backend.
But, the constants don't exist in either case
Is it safe for me to just comment out the couple lines it complains about, or should I actually map the constants?

[idle]

I updated the code in 1st post. it maps the linux errors to the #wsa constants but linux doesn't have
#WSA_IO_INCOMPLETE or #WSA_IO_PENDING equivalent so they are in the global scope

[Quin]

Thanks Idle, you're amazing! It's my bedtime, nearly 1 AM over here, but I'll try this tomorrow and let you know how it goes.

[Quin]

What's that they say about blind people?

We have great sleep schedules?

Wait no...definitely not that...


Instead of sleeping, I tested this, and it works, both with the ASM and C backends on Debian 12!
Thanks, Idle!

[Quin]

Wow, a tripple post...think that's a record for me. Sorry Fred!
I notice that when I run both the client and server on localhost, I can send packets of large length with no problem.
However, if I put the server on a remote server (a Linode) and run the client on my local machine, the maximum packet length is around 2750 bytes?
Any ideas?

[idle]

Isn't that like standard MTU maximum transmission unit on a lan? Seems to ring a bell

[Quin]

Isn't that like standard MTU maximum transmission unit on a lan? Seems to ring a bell

Sorry, no bell being rung here.
Are you trying to tell me that I literally just cannot send packets of greter than 2750 bytes on anything that's not my LAN? Frankly I don't believe that.

[miso]

It also depends on your internet connection type. ISP's also set MTU for their network, a PPOE protocol's default is just 1492. (ISP supporting RFC 4638 protocol can circumvent this.)

[Quin]

Sorry, but what?
Why should my ISP be able to limit my packet size to 1492 bytes, when I can go into a web browser, load up google.com, and get served hundreds of megabytes of javascript and CSS? That's just a TCP socket too, no?

[miso]

They stream it, and assemble together at the endpoint. And yes, ISP-s sets an MTU of their own.

[Quin]

They stream it, and assemble together at the endpoint. And yes, ISP-s sets an MTU of their own.

With all do respect to Idle, what are the point of these functions then? I expected to be able to send 10K+ chars over the wire exactly like I can over localhost. However, this seemingly only changed it from being able to send up to 700 or so bytes (a single read/write on the socket) to 2750 ,because of...some ISP limit? WTF? I hate networking.

[miso]

I don't know, how screen readers work, if they work on a command prompt. In windows, if you open CMD, and type

Code: Select all

netsh interface ipv4 show subinterfaces

it will display your computers own MTU values. The loopback pseudo interface 1 with the big number is the mtu value when you send data local to local.
The small ones will be the externals.

[Quin]

Oh my goodness, yeah, that is quite small. Around 1500 bytes.
This is unacceptable for my application. I'm building a realtime, peer-to-peer chat app, and users being limited to whatever their MTU vallue is for a packet length is super far from ideal.
Any suggestions?

[miso]

My knowledge is is limited, and I only work with UDP. I divide my big data and number them in their assembly order. I send them all one by one and assemble when all arrived. You can send big data that way. I use a header that is part of the packet. (TCP might circumvent this on its layer, I can't give reliable advice for that because my lack of knowledge with TCP)