Hello,
Keil support did not reply yet - but am I correct in assuming that RL-ARM is now a part of MDK, and that each user needs to have a RL-ARM license paid for separately in order to be able to use FlashFS/TCPNet etc. (many samples in MDK 4.20 are broken, but the one that I did manage to compile failed to link complaining that my license is insufficient - that did not happen with MDK 4.14 !) ? If so, this is a HUGE expense. If my boss asks for my opinion (and I think he will) - we're going open source!
A thing with Keil's ARM tools is that they cover everything from the smallest Cortex-M0 to the hugest ARM-9.
Maybe Keil need to consider a low-cost "lite" version for the low-end Cortex-Mx; eg, with microlib only, no floating point, etc...?
"Sources close to ARM" suggest that this thread has been noticed...
I don't think the solution is to cut the tools into huge number of sizes. That means that customers will have to look at their component choices based on their Keil support.
The main reason for profit is in the amount of support Keil has to give to each customer.
So maybe they should instead differentiate the support - sell cheaper licenses with no support except bug fixes. Each such license will be almost 100% profit, since they may be sold completely electronically - just costing some administration to register the customer, and the bandwidth for people to download the compiler and service updates.
That would mean that small companies can afford to start with Keil tools even for short series of cheaper products. And when they do get into a critical situation, they can decide if they want to upgrade to a more expensive license with support included, or if they want to pay a per-issue fee for getting help.
But the goal must be to get more people to jump on the ARM bandwagon - picking market shares from Pic, AVR, 8051, MPC430, ...
Going the other route - only selling to fortune 500 companies means there will be too few people using the tools and finding bugs. So there will be few customers paying huge fees to get low-quality stuff. PC compilers are so good because there are tens of thousands of users who tests everything in the tools and sends back good error reports. If user A can't write a good error report, then user B, C, or D will manage to write something that does help locate find and solve a problem.