[Info-vax] Microsoft: Alpha architecture responsible for poor Windows file compression

[Stephen Hoffman]

On 2016-11-02 13:58:38 +0000, Simon Clubley said:

> Do any Alpha architecture experts here know if this is the full story ?

Does it matter?  Microsoft has certainly given an interesting 
explanation and — all discussions of its veracity aside — the last 
Alpha systems with Windows support were Windows NT and those systems 
were most of twenty years ago.  Given Windows 2000 never shipped with 
Alpha support — just that beta — and the quick follow-on of Windows XP 
in 2001, Microsoft clearly hasn't deigned to implement and deploy 
better data compression and the associated migration in the ensuing 
years.   Or this wouldn't be getting posted in 2016.

Newer data compression algorithms inherently mean dealing with older 
schemes and the associated data migrations, if you're planning on 
adopting the associated improvements.   Or it means using what you have 
— a long tradition on OpenVMS, too — and deferring or even ignoring the 
associated improvements.

This situation is not at all an unusual case.   Memory compression and 
live no-reboot system upgrades would get to deal with changes to memory 
compression — yes, there are systems that compress physical memory, and 
good reasons to do that — and there are obvious similarities with 
retiring insecure versions and rolling out improved disk encryption 
support as well.  This is also akin to what led to the mess of RMS 
files within the morass of the OpenVMS system and cluster 
authentication and configuration database, too.   Working around some 
old limitations, around upgrades, and  being (usually administratively) 
blocked from completely removing the limitation (for reasons of costs 
and compatibility).

For the Register, this is page hits.  For Microsoft marketing, a 
discussion of compatibility.   For others among us, an indication of 
how trade-offs around application compatibility can increase costs and 
implementation difficulties, and can slow or even deny improvements to 
end-users.



-- 
Pure Personal Opinion | HoffmanLabs LLC