Filesharing Software → [Newsgroups] SSL Overhead

I see a forty percent throughput loss in upload speed using SSL, but negligible loss while downloading. Is that common when posting, using either an SSL enabled reader or Stunnel and Power Post? I expect ten percent, or possibly up to twenty percent overhead, but forty percent seems excessive. I've had similar experiences, using five different servers of three providers.

Partially educated guess: poor re-encode (to ASCII) practices in your client software and/or your CPU/RAM isn't up to the task.

To post to usenet your binary coded files (audio,video,rars,7zs) have to be re-encoded into wildly inefficient ASCII character strings because that's all usenet comprehends. This is done by your usenet client.

On an SSL connection your client must then also encrypt the now-bloated ASCII stream before handing it the transport layer. Both are CPU and cache intensive which puts them in direct competition for the same resources at the same time. If you are shy on RAM to use for cache then disc activity will get slammed and really bog things down. But simply poor client coding might explain what you're seeing.

I'm guessing the core modules on many usenet clients are old. Yours may offer data chunk and/or cache settings which allow you to optimize for SSL versus in-the-clear transmission but I doubt it.

Have you asked around in forums dedicated to usenet tools and/or consulted your usenet provider? I don't use usenet for anything but what it was intended for (simple text discussion) so can't offer any good forums for your issue. G'luck.

----- edit:
To address your question more specifically, it's not SSL overhead, per se. Secure connections are slower to establish but once the handshaking chit chat completes the transport is data type agnostic.
This wikipedia article has a column (SSL/TLS) showing which support encrypted streams. May want to trial some other clients.
»en.wikipedia.org/wiki/Co ··· sreaders

Thanks for the response! It's a little more than I was looking for, which is user experience, but it's very informative.
I get similar results using hardware with half the cores, and half the core speed, as well as twenty five percent of the memory of the PC I used in recent tests, so I don't think it's hardware related. Future tests will be with different SSL versions and fewer server connections. I don't post often , but when bored, I like to assess the clients and services I use by posting.