<@U4CP1J0QP> I was looking at this issue <https://github.com/ktorio/ktor/issues/654> More or less I'...

# ktor

[Edoardo Luppi]

@e5l I was looking at this issue https://github.com/ktorio/ktor/issues/654
More or less I'm having the same problem.

In an old implementation of the TCP protocol I'm re-implementing they were using Java's Socket.getInputStream().available() to know the available bytes in the read buffer.
That buffer is 65k bytes, which more or less means I need a couple of iterations to read the entire server response.
I know it's a weird way to implement a protocol, but I can't do anything about it now.

With Ktor I find myself reading 4k bytes at a time (availableForRead), which slows it down quite a bit, also because the buffer doesn't seem to be immediately refilled.
Is there a way I can use an arbitrary size for that buffer? I'm not sure your answer there applies here.

[Edoardo Luppi]

I think I can workaround this by changing how I read data from the server.

I've seen ByteReadChannel has a readUTF8Line method. Which is mostly what I'd need.
However my protocol doesn't operate using UTF-8, but using charsets like CP037 or CP1047.

Could I implement my own readLine method - that doesn't decode the byte stream - using utilities that are already present?

[e5l]

Hey @Edoardo Luppi, thanks for the report. We're working on migrating to the new kotlinx-io library. After migration the issue should be fixed

[Edoardo Luppi]

No problem @e5l, I've built my own workaround for now and it works ok!
I really like kotlinx-io stuff btw, hopefully it will include much more utilities going forward

[Edoardo Luppi]

My use case is a custom TCP protocol with multiplatform support