C
cquirke (MVP Win9x)
Surface area - that's a great analogy!
My beat is the opposite; "self-aministering" home and small business
users, where one would like to restrict as little as possible but make
it easier to practice "safe hex".
What I like to do is generalize up from specifics to see if there's a
theory-level lesson to be learned - then apply that theory to new
situations and predict problems before they occur.
That's what theory is for, in this sense; to predict the bounds of the
possible... many things that one would like to believe are impossible
are merely difficult, or not even that; simply an opportunity
no-one's grabbed yet.
Great directions, but context oversnip
When I see something that bugs me, I ask myself; is it due to an
oversight, or by design? Part of assessing that is; what does the
vendor stand to gain/lose?
From MS's perspective, the wildcard in the pack isn't some new
competitor running off with the market - that can be spotted some ways
off. Instead, it's something that kicks over the legal antheap and
attracts public as well as legal attention.
Even a reluctant, corporation-sympathetic government may have to act
if the public is watching and prodding them to do something, and there
comes a point when the impact on big business in general may cause
them to Brutus in order to cut their losses.
That's why I see prudent modular design as being as much in MS's
interest as anyone else's, even if this misses opportunities to
"embrace and extend" into other markets.
Hey, they can read me for free
There's a triangle of capital, worker and consumer that causes
business to work best when these forces are balanced in power. This
facilitates the selection pressure that is often mooted as the most
efficient way to progress.
No choice, no selection; that's the problem as most of us see it.
No pressure, no selection; that's the far larger problem.
In the industrial revolution, a situation arose where a small number
of employers made it impossible for a large number of individual
workers to apply selection pressure - because a small number of
players can act in unison as a cabal. Organisation of workers into
trade unions redressed matters somewhat, and has brought social
benefits beyond the obvious (i.e. better wages for workers). For
example, union pressure can highlight public safety issues.
In the information revolution, a situation is arising where a small
number of vendors make it impossible for a large number of individual
consumers to apply selection pressure.
Right now, the software industry has been left to write its own rules,
and the result is a disaster for users:
- long-term patents, then product is destroyed and useage denied
- limited license rights
- sanction on self-repair ("reverse engineering" etc.)
- no vendor safety liability whatsoever
- EUL"A"s that are drawn up unilaterally by vendor
- EULA's hidden inside no-return-if-opened boxes
- moving goalposts for expectations of service
- forced ongoing relationship with vendor
- vendor reserves right to unilaterally poison-pill user's system
- user to obtain "product recall" replacement code at own cost
- user to obtain "product recall" replacement code at own risk
- in some cases, proprietary lock-in where data is concerned
- in some cases, version lock-in and lock-out to force upgrades
Organizing the users is difficult because the whole shebang is likely
to get hi-jacked by the industry - whether it be pro-vendor shills or
(more likely) an agregate of "everyone except Microsoft".
And believe me, being dragged back to the mainframe era (now disguised
as "thin clients" renting time on application servers) will re-invent
the need for a MS to promote "personal computing".
DOS business and consumer PCs were pretty well immune to that sort of
problem, as they hadn't gained their WAN wings yet
I think there's deffo an attention-defecit disorder there, i.e. a
short attention span. Things are designed in a certain way for
particular reasons, then those reasons are forgotton (seing as the
design prevented those problems from happening) and the clue that
informed the original design gets tossed overboard for some trivial
"shiny things!" sort of reason.
For example, consider the matter of WinME's automatic file system
cleanup after a bad exit. The original reason why this facility was
deigned was so that by the time Windows starts to write to the file
system, the file system would normally have been repaired so that its
now safe to do so. That's why the Win9x versions that had the feature
had always done this using real-mode Scandisk, because that could
complete the job before Windows starts writing to disk.
But because WinME has been tinkered with to make at look as if it's
less connected to "DOS" than earlier Win9x, it now runs this automatic
Scandisk while Windows is loading (and writing to the file system).
It's like making jumbo jet tyres out of glass because it looks so kewl
when the sun shines through them on take-off, forgetting why jumbo
jets needed landing wheels in the first place.
Yes, there's been a rather unfortunate NIH problem in effect. MS
likes to pretend that the areas they extend into had never existed
before they came along, inventing everything from scratch.
I can understand the joy of sweeping with a new broom, but part of
that joy is not having to make the same mistakes.
An old design may become unwieldy because of a lifetime of corrections
necessitated by the "school of hard knocks"; designing afresh should
take those lessons on board and result in a smoother design that works
as well or better. Re-inventing the *original* design and then having
blunder through all the same mistakes from scratch is duuumb; at best
it will result in a scarred but stable version years down the track.
Mind you, some ideas are so dumb that none of the pre-existing
solutions ever made those mistakes - so there's no old newsreels of
fiery crash-and-burn to watch and learn *those* lessons
Lovesan penetration + CIH payload = economic metdown.
The equation is that simple... imagine the business impact of not just
one tower full of offices, but most of everyone's infrastructure (no
matter how widely distributed) not only knocked offline, but blown
away at the hardware level (think soldered-on BIOSs with no fallback,
think proprietary laptop motherboards, think old motherboards tyhat
can't be bought off-the-peg anymore).
The infosphere is indeed a selection-pressure environment, just like
the biosphere, with one difference; mutations (and indeed software in
general) are made, they do not as yet spontaneously arise.
I've explored those themes recently in alt.comp.virus; a google on
"infosphere" and "cquirke" will prolly find them.
It's certainly complex enough to defy deteriminism, and become
interesting as a result. In fact, that's what interests me the most.
Just as you can model natural complexity in a computer, so you can
look at the most complex biosphere survivors (and see how these differ
from simple ones) and learn lessons.
For example; bacteria have no nucleus membrane, and the genetic
material just lies around, not really organised into chromasomes.
Anything can squirt in new code, and just about everything does;
bacteriophages, conjugation with other bacteria that aren't even the
same species, human researches wanting to trick E.coli into
manufacturing harvestable human Insulin, etc.
Multicellular organisms have a clearly-defined way in which code is
exchanged, so they still benefit from breadth of variation as do
bacteria, but it blocks out "dropper attacks". Wrong species; don't
bother to apply. Right species, but looks dodgy? Forget it! Food
that is eaten is smashed down into fragments too tiny to pass code
unchanged, and only then does it enter the system.
Humans also have enough on-board and retrospective intelligence to
predict problems and fix them, but alas, the sealed nature of the cell
and nucleus makes it hard to apply direct maintenance. You may know
that certain protiens are part of your own body, but you can't stop an
inappropriate Rheumatic Fever immune reaction from shredding your
heart valves. A hermetically-sealed, DCMA-enshrined future Windows
could become exactly that kind of death-trap.
I have a hunch MS may just re-invent the entire wheel, and use
whatever by-then glaring deficiencies there will be with our current
Windows as reasons to upgrade. Except this time, you may never ne
allowed to leave the shop with your purchase, and will be charged rent
Two lessons to bear in mind, on such a strategy:
1) Every layer is only as good as one it rests on
2) No code is ever perfect
3) Don't even let the system override the owner's control
The question is, who is the owner? The corporate administrator? For
consumer PCs, the user via a simplified UI - or MS as the default
system adminstrator of the world? Or none of the above; instead, a
global cabal of media pimps and cronies?
Yep - and it's something that has to be designed in. Not in the form
of "in passwords we trust"; in simply making hostile actions
impossible. That agenda may be best served in the corporate world via
NT's existing model, but in the consumer world, users are already
familiar with one better suited to their needs; physical access.
Windows is learning its ABCs. For example, it now knows that data is
best located on a per-user basis, rather than dumped within each
program's own code files. It attempts to select system scope while
ignoring personal data scope; that's what SR tries to do.
What it still doesn't have a clue about is the concept of "suspect
files" or risk. It gaily dumps anything IE downloads into "My
Documents" and nests Messenger's "My Recieved Files" thetre as well -
the concept of "data hygiene" has yet to glimmer.
Now I am not for a moment suggesting attempts to track the origen
context of a file as it passes through the system - that's what IE's
security zone attempts to do, for example. But just as I'd never
consider IE's zones to be even sand-tight, it's still a useful net.
Such awareness only becomes counter-productive when you consider it to
be air-tight enough to confer all sorts of ill-advised rights to those
"inside". As long as treated humbly, a "suspect zone" is useful.
That sort od scope awareness would massively simplify maintenance, and
that's against the interests of the software industry. Not because
they want you to have problems, but because they don't want any
scrape-off of thier programs to be as functional as a formal install.
The nesting logic is asinine, IMO - why the hell would you want
unsolicited incoming malware and your entire .MP3, videos and picture
collection nested within your data set? How is that going to fit on
limited-capacity backup media? How safe is that going to be to
restore, after a delayed malware payload that also clears forensics?
It's the right idea, but it's too difficult to relocate these around
(TweakUI helps) and - fatally - you cannot *pre*load where these
things will be located when new accounts are created.
Unless using roaming profiles, the relevant registry part is located
within the user account subtree AFAIK.
The trouble is, the safety scopes are too mixed up. For example, any
shortcut collection is a malware dropper opportunity, with StartUp as
the big prize, and the same applies to registry content. That's why
you can't write-share the whole of "Documents and Settings", or back
up the whole of this subtree with safe-to-restore assurance.
My own practice is to locate small data on a small volume on it's own,
and automate a backup of this to another small volume as an unattended
Task. Such backups can be pulled onto other LAN peers to create a
multi-redundant "holographic" store that resists all but total
infrastructure destruction, even if no-one ever inserts a backup
medium. Temporal depth can be created using the same multi-shot
cyclical backup logic that Win98's RB00?.cab uses.
Data too big to automate in this way is stored elsewhere, as are
pictures, video, music and other collectables that, while you'd want
to see again, are not as irreplaceable as your own data.
Suspect files (emaul attackments, downloads, peer-to-peer file sharing
bins, etc.) go in their own subtree, under the muzzle of a battery of
on-demand scanners that can be launched via a single QuickLaunch click
This works well, until some idiot creates new user accounts. Then
everything duhfaults to being in C:, which slows down as those
absurdly huge per-user IE web caches fill up.
Yep. MS seens incapable of maintaining any sort of continuity there;
that attention defecit problem again?
Data may be in; "C:\My Documents", %WinDir%\Personal,
SomeProfilePath\AsAbove, Local Settings\some\path, who knows? Why not
bury things in the bowels of the OS subtree? Duh.
The worst was when Office 97 unilaterally imposed "My Documents" as a
new "system folder" on the OS (since when does an *application* have
that authority?) without any UI other than RegEdit to manage it.
Unforgiveable, and a good Exhibit A why splitting MS into OS and
application companies would be of some benefit.
I mentioned attempted pan-directory scope awareness earlier on, in
connection with System Restore. FAST attempts to do the reverse;
fillet out everything that *isn't* "system" and present this to you as
a restorable backup. Gary Woodruff clued me into this, and he has a
good page on the topic somewhere Googleable.
Both SR and FAST are version 1.1 and 1.0 (respectively) attempts at a
"mission impossible" brief, which rears fears of vendor hubris.
AFAIK, both are equally closed "trust us" systems, in that I know you
can't pick-and-choose amoungst SR's stores (you can't even clean
malware there <g> ) and this may apply to FAST too.
Their value may not be in whether they work or not, but that a scope
awareness may better inform decisions on where to put things.
However, MS doesn't share the UNIX view that the file system directory
structure should also define other scopes (such as user rights, etc.).
I'm in two minds about this; frankly, a single structure can turn out
to be a straight jacket (hence Lotus Improv and Excel pivot tables -
and look how easy to use they are!). Scopes vary and overlap, e.g.
- access speed are file system risk exposure (which volume?)
- per-user vs. system
- data / large-data / system
- safe / unsafe-infectable-risky
- in-system material vs. stuff coming in from outside
- shared / read-only-shared / not shared
One could conjure up an elaborate system to manage all of these
many-to-many relationships, but it's all sandcastles in the air if
anything goes wrong within the levels beneath (malware intrusion,
flaky hardware that sees everyting as raw sectors).
There has to be a way "under" this to maintain it, else you have a
data death-trap on your hands... an elevators-only skyscraper with no
windows or fire escape. Be careful what you wish for.
Who has the rights to "maintain" the system at that level? The owner,
of course. In consumerland, that should be a matter of who has
physical access; if you want a virtualized but nearly-as-safe model to
facilitate corporate remote admin, pay up for the Pro version.
If OTOH this "ownership" is taken out of the hands of the person
buying the product... well, maybe it's time to storm the Bastille.
You can't tune into a cabled LAN unless you have a pre-existing
presence there. You can't attack a subsystem through a port that
doesn't exist. You can't hide under the system's own access
protection if the system doesn't deny the owner full access to
anything. You can't get your code auto-interpreted if the interpreter
not only doesn't interpret script, but doesn't have "unchecked
buffers" (hint; StrCopy is not your friend). You can't break through
password protection to use a service that does not exist. There's no
point in crossing user account rights and security zones if the
functionality you seek does not exist.
There's a clue meme in there, looking for some frontals to infect
The rights you save may be your own