NSD History

Revision

Version 1.4 (1997-05-15)

History & Conformance

This chapter will tell you about changes in different versions of NSD and about the reasons behind the changes.

The most recent changes are at the top of the list:

   - Public Revision 1.6: Minor text cleanups and clarifications
     for publication on ftp.amiga.de. No substantial changes.

   - This version of NSD fixes an obvious typo in the
     define list that nobody noticed or reported previously.

     The correct define name for audio type devices is:

       #define NSDEVTYPE_AUDIO         8   /* like audio.device */

     It used to be documented as NSDEVTYPE_AUDIOARD, which is
     wrong.

     The text "large enough" describing the mn_Length check has
     been replaced with a description of "at least the minimum
     required size". It had turned out that people misunderstood
     "large enough" as meaning "exactly of the required size",
     which is obviously wrong in the context of devices supporting
     multiple request sizes for different types of commands.

     Similar to the change mentioned above, other places have been
     reworded to avoid misunderstandings.

     The basic NSD command set description reflects the
     distinction between command set and semantic consequences
     now by elaborating on the device types. There is also a
     description now on the standard V40 command slots for the NSD
     device types.

     The NSDEVTYPE_TRACKDISK specific section has been extended
     with a description for ETD like commands. Also, a note has
     been added that e.g. the V40 cd.device as used by C= for the
     CD32 game system should not be forgotten. It is a C= V40 OS
     device "like trackdisk", too. General clarification comments
     have been added in various places.

     Notes for NSDEVTYPE_SANA2 have been added.

     As it turns out, NSD is not only about identifying device
     capabilities, but also about safe use of a device as a
     consequence. While 100% safety can probably never be reached
     within the existing frame, there are simple ways to make
     usage of devices a lot safer. The text now reflects this.

     Due to the complexity of the issues involved, the document
     has been split into sub documents. The wording has been
     changed in several places to reflect this.

     Additionally, a "Thoughts & Consequences" section, a "Future
     Directions" section, and this "History & Conformance" section
     have been added as sub documents.

     The requirement of changing TD_GETDRIVETYPE for NSD
     trackdisk like devices has been removed. The DRIVE_NEWSTYLE
     identifier for trackdisk like devices doesn't serve any
     useful purpose. It never has and has been introduced only
     because of an oversight during the revision process. While
     the trackdisk TD_GETDRIVETYPE command generally isn't exactly
     the most useful command in the world currently, and while it
     has been documented that anyone using a new style device
     should use TD_GETGEOMETRY, redefinition of TD_GETDRIVETYPE is
     obviously not a useful thing in this context and it would mess
     up the implementation of a true NSD trackdisk.device
     replacement.

     The NSD recognition sequence contains an additional simple
     check now. Correctly written software will continue to work
     well, and software written now will be safer.

     A recommendation has been added on how to set up a reused
     request structure for another OpenDevice() to do a general
     NSD query. As some strange device expect parameter data being
     set up on an OpenDevice() call, the request structure should
     not contain or point to garbage data.

   - Revision 1.4 of NSD got published soon after the release of
     the Amiga Developer CD 1.1. It added the requirement to check
     mn_Length in OpenDevice for at least the minimum required
     size to be handled by the device and documented the "safer
     assumption" for querying a device. It had turned out that
     SANA devices can barf on a general query. As a device
     implementing the mn_Length check will at worst fail to open
     instead of messing up the system which could have happened
     previously, NSD got safer and better this way. A client not
     using the safer assumption is more likely to fail with bad
     devices, but it will still work with the updated NSD spec if
     it doesn't have bugs.

   - Revision 1.0 of NSD, as published on the Amiga Developer CD
     1.1, unfortunately had no RCS Id string. It should
     be referred to as NSD revision 1.0.