> On Wed, 21 Nov 2007 20:19:28 -0800 > Greg KH <greg at kroah.com> wrote: > > > > > One funny thing, a number of companies have contacted us already, asking > > for Linux support for their devices, only for us to tell them that all > > of their devices are already supported just fine, no new work is > > needed. They were pleasantly supprised :) > > > > I think such comments should be made with caution. In most cases where > the vendor hasn't been involved, there has been some level of reverse engineering involved. So the driver can be based on flawed assumptions. Also, there's the issue of erratas. In short, just because we have a driver doesn't mean we don't want vendors to get involved. > > Rgds > -- > -- Pierre Ossman > The biggest problem I've found with development in which the vendor was not involved lies in error detection and recovery. Every vendor has a different policy. The most robust code requires a lot of development time and $$$, so many OEMs elect a scaled-down approach. The trivial example is a paper-out fault on a printer. This happens frequently, and all devices and processes should anticipate it and handle it correctly. However, there are many other faults which can be created only with great difficulty, if at all. These represent a serious challenge to the reverse engineering paradigm. Creating a complete and robust support module for a product which is working properly is only a small part of the effort. The balance may not be possible without the OEM's involvement. BR Leigh
