On Tue, 08 Aug 2006 17:54:17 GMT John Navas wrote: | On Tue, 08 Aug 2006 10:00:06 -0700, Jeff Liebermann | wrote in | : | |> snipped-for-privacy@ipal.net hath wroth: |>
|>>The hardware savings exceeded the OS licensing cost, so Cisco |>>makes more money this way. |>
|>I beg to differ. The only major difference between v4 and v5 hardware |>is the size of the flash RAM and main RAM. I've compared the boards |>in detail. All the same parts from the same vendor. My guess is |>perhaps $0.50 hardware savings per board in large quantities. Let's |>see... Hynix 16Mbit SDRAM chips are about $0.70/ea in 100K quantites. |>What hardware savings? |>
|>Last time I checked, the VxWorks DevSys is $50,000. My guess is that |>runtime licenses are about $5 per unit in 10K piece quantity. Probably |>considerably less in millions[1], but methinks it's still more than |>the paltry parts cost savings. Considering that memory chip prices |>are in decline (after last years increases), while software license |>fees seem to be increasing, it's also a lousy long term strategy. |>There's even allegations of price fixing in the memory market: |>|
formatting link
|>
|>See: |>|
formatting link
(900K) |>for a comparison photo. V4 on the left, V5 on the right. The |>components have moved somewhat but they're basically the same. |>
|>[1] Duz anyone know the large quantity VxWorks licence costs? | | I personally think the change probably had more to do with efficiency, | stability and support. With all due respect to Linux advocates, it's | not a terribly good embedded real-time OS. It was probably used in the | first place as a matter of expediency and cheapness (and coolness), not | suitability. VxWorks is an excellent platform. Likewise QNX (my | personal favorite in this area).
So you think that the increase in support issues I've read about with the version 5 was just due to the newness of VxWorks for the company, and that in time this will work out?
Personally, I've have probably gone with NetBSD for such a thing (and yes, I've a Linux advocate first, BSD second). I may still try to port NetBSD over to WRT54GL some day. But I'm going to do Linux first because I am already familiar with hack it even at the kernel level. I know nothing about VxWorks internals, so I can't really say if it's good or bad.
I do know one issue with Linux (but not NetBSD) is that any changes made to the internals of Linux must be included in source releases. That may have been a motive, too. Maybe they had a new feature in mind but did not want to release its source code.
I don't yet know how much Linksys changed the Linux guts to work for the WRT54G. I might look at their source if I have time. But as soon as the current project ends, I'm definitely going to be grabbing something like OpenWRT for the WRT54GL and hack from there.