TODO

TO BE DONE in LibDSK 1.5.x:

- [Apple] Improve resource fork / MacBinary support in DC42 driver.
- [QRST5] Reverse decompression of QRST5 file.

- [Apple] Handle at least basic support of 400k and 800k MFS / HFS filesystems


FUTURE LibDSK 1.9.x:

- Redefine DSK_GEOMETRY (breaking change):


	typedef struct dsk_zone
	{
		dsk_pcyl_t	dz_cylfrom;	/* Cylinder / head ranges */
		dsk_pcyl_t	dz_cylto;	/* Zone applies if */
		dsk_phead_t	dz_headfrom;	/* cyl / head fall within */
		dsk_phead_t	dz_headto;	/* the range */

		dsk_psect_t	dz_sectors;
		dsk_psect_t	dz_secbase;
		size_t		dz_secsize;
		dsk_rate_t	dz_datarate;
		dsk_gap_t	dz_rwgap;
		dsk_gap_t	dz_fmtgap;
		dsk_recmode_t	dz_recmode;	
		dsk_flags_t	dz_flags;	/* Formerly part of recmode */
		int		dz_nomulti;
		int		dz_noskip;	
	} DSK_ZONE;
	

	typedef struct struct dsk_geometry
	{
		dsk_sides_t	dg_sidedness;
		dsk_pcyl_t	dg_cylinders;
		dsk_phead_t	dg_heads;
		int		dg_zonecount;
		DSK_ZONE	dg_zone[1];
	} DSK_GEOMETRY;

	will need:
		dsk_geometry_alloc(int zones);
		dsk_geometry_clone()
		Serialisation format for GEO2 block in LDBS

	new ini file format:

	[Type]
	Description = foo
	Sidedness = Alt
	Cylinders = 40
	Heads = 2
	Zones = 2
	Sectors[0] = 8	# Sectors[0]= abbreviates to Sectors=
	Sectors[1] = 9

	etc.

	


- [CopyQM] Use real CopyQM on real hardware to image some DOS-format
  floppies so I can see the BPB.

- [CPCEMU] Debug the new LDBS-style driver:
	* Need to try on a few more test cases. extextdsk tests all seem
	 to pass; would be nice to try some old-style DSKs too.


- The cpcemu driver contains a chunk of code to deal with multitrack reads
 and writes. Write some test cases to exercise it, and check that the LDBS 
 common driver behaves the same way.

- Partition support.
	It looks as if the way to do it is to have a filter driver.
	Something like 
		dsk_err_t dsk_part_open(DSK_PDRIVER *guest,
					DSK_PDRIVER host, 
					DSK_GEOMETRY *dg,
					int npartition);

	This requires:
		- Reference-counting on the host. Either fail dsk_close()
		 if its refcount is > 1, or let it succeed silently.
	
	Once dsk_part_open() is in place, we can also allow partitions to
	be passed as filenames:

		partition:n,filename

	which will have a dsk_open that:

		* passes filename to inner dsk_open
		* do a dsk_getgeom
		* pass the resulting parameters to dsk_part_open

	Also: False partition (ie, add a partition table to an image)

		dsk_err_t dsk_false_partition(DSK_PDRIVER *guest,
					DSK_PDRIVER host,
					DSK_GEOMETRY *dg,
					unsigned char type);

		with matching open-by-filename:

		falsepart:n,filename

	The alternative to a filter driver is dsk_option_set with a variable
	of PARTITION, but that has certain problems (eg: you can't have
	multiple partitions open at one time)

	[nb: What sort of partition table do the likes of +3e use, because
	 that's going to give all sorts more fun]

* Weak-sector support is new and may be buggy.
* Write more functions for dsktool.
* The INT25/INT26 driver doesn't seem to work in a DJGPP compile; it fails
 with error 0x701F.
* The INT25/INT26 driver keeps spitting out diagnostic messages.

* dskid doesn't work on a DFS .DSK file. Why? Because get_id doesn't fail?
* Support for DMK file format.
* Java support for the Reporter classes is buggy [seems to work now 11-5-2003]
* Plugin architecture:
	dsk_register_driver
	dsk_register_compression
	dsk_register_remote

* Support remoting libdsk...
	>> May need to write a PCW backend that dances on the controller
	  rather than using decorum and the XBIOS.
	>> May need to write a Win16 version of serslave.
	>> May need to write a Win16 RPC client.

* CopyQM: Compare to real BBC Master 512 disks 2-4.