Regulars will have seen my topic posted early today.
http://www.xcore.com/forum/viewtopic.php?f=8&t=820
I mention it, because it's very relevant to this current thread.
At the moment, I'm very concerned that XMOS is confusing image with substance. I say this after returning from a break working on a different project, and finding a few improvements (XDE updates mainly, better modularity) but a solidly closed front to the community in the areas that really matter: Code and Tickets.
Meanwhile we have two wonderful web sites, that have cost considerable investment. Investment that may have been better spent elsewhere. Feel free to differ on that aspect, but let me make the important point...
If I am to bind my future and the success of my clients to XMOS devices (and now that this is
software defined silicon, by extension, XMOS code) then I need XMOS to grow up a little.
I have spent three days working intensively on xtcp, to convince myself that earlier statements by other respected members of this forum are absolutely true. I now fully accept that xtcp is unusable for anything but the most trivial purpose. I find this disgusting, considering that the code has been hanging around on this site for months, it has seen three minor versions, and there have been comments on this forum about its quality in the past.
Looking on the bright side, my three day investment has gained me some real knowledge, and makes me very qualified to turn my back on the whole stack until I see some signs of intelligent life from XMOS.
Intelligent life?
Yes!
Sign #1: Let's see a
public tracking entry that acknowledges that packets are dropped all over the place in xtcp, that it's very timing-sensitive, and that it blocks completely after random periods (seconds to a minute). This will save some customers' valuable time, and it will be a place to gather more information.
Sign #2: Let's see the
entire source code for the stack moved to github, or similar. (XMOS: Take this message as an offer to setup Subversion, if you prefer. Or go to Netsight in Bristol who will do a wonderful job for you.)
Sign #3: Let's have a public acknowledgement that XMOS will be taking community contributions to that code.
Sign #4: Let's see a representative test suite developed as part of the whole network stack. I expect that suite NOT to work at first, but see the next point -- help is at hand
Sign #5: Let's see step-by-step changes in that codebase, that the community can test and comment on. A true sign of life.
Sign #6: Accept contributions from outside. Work with the community to define the best way forward. We will know this when we see it.
Then continue to release the occasional zip bundle, but keep the core code moving along in the repository, and be proud of the tests that prove XMOS has something to offer.
With apologies, my XMOS haiku of the day....
Code: Select all
I close XDE
XTCP Frustration
Tomorrow brings change
Best wishes,
--r