Re: IPac driver organisation

[Andrew Johnson <anj@aps.anl.gov>]

Hi Doug,

Doug Murray wrote:
> Hi Andrew, Iâve noticed the IPac directory contains support for a number 
> of carrier cards.  Would it be possible to add drivers for:
> xycom9660
> acromag9670
> hytec8002
>
> We use these at SLAC, and weâre using RTEMS so weâve had to OSI-ify 
> these packages.

Andy Foster wrote the Xycom/Acromag carrier driver which is licensed 
like the rest of Ipac under the LGPL (currently V2.1; I may move Ipac to 
V3 before the next release).  If Andy is happy for it to be included I'm 
willing to add it (with a few minor changes) to the Ipac distribution. 
The next release of Ipac already uses the OSI interfaces for everything 
except the tyGSOctal driver and ATC-40 carrier driver, so I'd want an 
OSIfied version rather than the vxWorks only code that is in his 
xycomIpac-1.0 release.  If you have an OSIfied version, please send me a 
copy.

The Hytec driver is Copyright Hytec, and the source code we have here 
does not mention any licensing terms at all, so I can't distribute it 
anyway.  I would not be willing to maintain this code at present, which 
also provides some VME64X hot swap capabilities that I do not claim to 
comprehend.

> Also, I see drivers for a few IP cards in that directory, such as CANbus 
> and an Octal UART card.  Is the idea that new IP drivers be placed in 
> this directory as well?

No; those module drivers are only still there because I maintain them 
and I haven't been sufficiently motivated yet to remove them to their 
own support module(s).  I don't want to add other IP module drivers to 
drvIpac because I don't want to take ownership of anyone else's code.

The EPICS build system is designed to make modules easy to create and 
use; just glomming a whole lot of unrelated drivers together into a 
single support module makes maintenance take longer and discourages 
people from releasing new versions of the resulting mess (the private 
'share' module at APS for an excellent example of that particular 
nightmare - 7 record types and 29 device supports that nobody wants to 
own or touch if they can avoid it).

> On a related note, the GPIB IP card (IP488) is located in the Asyn 
> directory.  Should IP modules be put under Asyn?

Since that card is supported by the devGpib subsystem which is currently 
a part of Asyn, I think that is the appropriate place for it.

IP module drivers should be released either individually or in small 
groups of related drivers, making it as easy to put together the 
software for an IP-based IOC as it is for the hardware.

- Andrew
--
The right to be heard does not automatically include
the right to be taken seriously. -- Hubert H. Humphrey