[sclug] Motherboards and SATA

Roland Turner raz.fpyht.bet.hx at raz.cx
Tue Apr 6 08:48:49 UTC 2004

Alex wrote:

> The usual argument is one of a) they don't want to support all the different
> Linux distros,

A clear furphy; make source available allows _others_ to take on this burden, and for Adaptec's and nVidia's gear in particular, people are standing in line to do this.

> b) revealing the source of the driver will reveal some novel
> design aspects of the hardware or

This is the argument that I have heard most often and is essentially addiction to a lie. I can't quite put this sentence together in a way that makes sense, but the lies are roughly:

- our engineers are far and away the best in the world and therefore our products contain valuable innovations that no-one else can even think of, but could copy if we showed them and, because we're too lazy to apply for patents we therefore think we should keep them secret by not publishing our source code,

- a determined reverse engineering effort would not successfully extract these magnificent innovations from our object code, because our competitors' engineers being far less intelligent than ours cannot operate a disassembler, nor analyse algorithms

- _we_ don't do reverse engineering (but please don't look to closely at the skunkworks in a city near-but-not-at our HQ), therefore our competitors must not, therefore withholding source code is an effective way of keeping secrets

- even if someone could and did reverse engineer our products, it would take so much time and effort that it would not be worth their while, so withholding source puts up an effective competitive barrier

This last point despite my guess that all hardware manufacturers in a particular space (disk controller, NIC, graphics adapter) do have reverse engineering labs and do have, within weeks of release, a detailed knowledge of the innards of any competitor's product.

Further, Eric Raymond suggests that having competitors copy your released products is exactly what you _want_ them to do, because it relegates them to competing in the also-ran space, not the market leader space. There are some holes in this, but an element of truth I suspect.

I had always assumed that the real (but unstated) reasons include those relating to the code itself:

- most corporations are more than a little ashamed of their source code; witness any major release of code to open source of a formerly closed product; months are spent "sanitising" the code to remove any evidence of a sense of humour on the part of any programmer. More practically, removing potentially offensive comments probably is a real issue.

- related; often closed source code isn't solely the product of the apparent vendor. Frequently, third party chunks are licensed, but cannot be opened by the apparent vendor. Opening the source that they could open would not in fact yield a working product (consider Mozilla, which had exactly this problem).

But here is one that I simply had not considered:

> c) differentiation between different
> models at different price points is purely a driver issue, and revealing the
> source will allow purchasers of the "model 100" to get the performance of
> the "model 500" simply by hacking the driver's detection routines.

While purchasers of software are accustomed to paying widely varying prices for what amounts to a CD because of the value that they feel that they can extract from use of the embodied application (people will pay for more for the license to use the software on a CD containing SAP R/3 than for the equivalent right with respect to a CD containing Windows), people buying high-end hardware are typically being led to believe that the higher-end device attracts a higher cost of manufacture, not simply that more of the vendor's software R&D costs are being amortised against the purchase. This sails pretty close to fraud, and yes, represents a broken business model.

- Raz

More information about the Sclug mailing list