Who needs a specification?
Published: 2008-04-25 13:04:42
I am adding a debugging function to the software I am working on. In it, I am sending a command to enable a specific feature in the drive the software talks to. I am doing this in accordance with the latest version of the specification, as published by T10, the technical committee overseeing various SCSI protocols. I am sending the same command to drives from three different brands.
- On one brand (let us call it the blue brand), the command succeeds, and the feature I am enabling does indeed become enabled, and is then accessible.
- On another brand (let us call it the black brand), the command fails with an error code, pointing to a completely different byte of the SCSI command I am issuing. A byte that is not at fault, as far as I can tell after inspecting the byte-stream sent.
- On the last brand (let us call it the cyan brand), the command succeeds all right, but the feature I am enabling does not become accessible. The drive continues to report that the feature is enabled.
Well, one out of three reading the specification and actually following it means 33 %. Not too bad, I guess.
Now, where have I heard of that kind of lack of support for standards before? I guess it is naïve to think that that kind of thinking is somehow limited to the browser business, but that this occurs everywhere when there are people reading standards documents and trying to implement what they describe. Especially when the standard is rapidly evolving.
Tags: interoperability scsi standards t10
This was originally posted on My Opera at
http://my.opera.com/nafmo/blog/show.dml/1975279
Please note that links may be outdated and any information included here
may be obsolete.
← You know you are doing low-level software development when… | Download Jack Rubinacci’s latest album for free (as in beer) → | Back to the post index | Back to the archive index | Peter's homepage