From: "J.C. Wren" <[email protected]> (I agree about the joys of C-Spy and other IAR tools.)
> One day I shall be interested in hearing why TI (and Atmel, for that > matter) think that protecting parts of the JTAG command structure is a > useful idea. I believe in IP, and I believe there are some things that > users don't need to know about. But the fact they're willing to share that > information with some parties and not others is disturbing. It seems to > indicate a managerial structure that doesn't really understand what a user > wants, but rather a structure based on unfounded fears. Do they really > think some is going to clone the core because someone knows how to read a > register with JTAG? And if they aren't going to tell, the least they could > do is say why they think they shouldn't. > > --John > (Note - these are my thoughts - I haven't asked the companies themselves.) I think the reason manufacturers keep tight lids on things like jtag details is quite simply so that not many people get to know the details (stating the obvious, I know). If we ignore the whole concept of open source software for a minute, then who is actually interested in knowing these details (I mean, who is going to make use of them rather than just looking at them because they are the sort of person who likes to know *everything*)? Almost entirely, it is people wanting to make hardware debugging or programming tools. Anyone serious about this can get the information under NDA. But the manufacturers want to control this for two reasons that I can see - it makes it far easier to deal with changes to the specs (they need only tell a few people, rather than leaving the world full of incompatibile tools with no software updates), and it avoids the situation where anyone can build cheap but not entirely reliable tools. If a customer buys a programmer which does not implement flash programming entirely correctly, so that the target devices die after three programmings, they will blame the chip and not the programmer. The only problem with this is that with open source software, you are inherintly breaking NDAs (as an alternative, maybe you could write the proprietry stuff in tightly-coded perl - it would then be open source, yet illegible to humans :-). I think the current solution is a good one, at least as good as is possible without freeing the proprietry information. As much as possible is open source, which has two big advantages - the code can be reused (for example, the avrgcc folks could use a similar tactic starting from your rproxy), and we users (most of whom know at least a little C) can be far more useful for helping to debug and enhance the programs. Anyway, many thanks for the work you've done so far. David
