Well, I'll be a monkey's uncle. The Amazon performance problem seems to be fixed.
After months of muttering to myself about the problems with Amazon, it appears to be fixed. RIght now I have a half dozen Amazon tabs open in my Firefox browser (on a well equipped desktop PC with Debian) and performance is good.
At the same time, I notice that when I hover my mouse over the "My Cart" icon, upper right of most pages, I no longer see a list magically appear of what I have in my shopping cart. I have to click on the Cart to go there.
This may well be the sort of reasonable tradeoff that was required to get adequate performance - tune down the glitz a little.
This was exactly the reason that gave the final push to ditch Firefox for me as well. Seriously, how can a page that's seen by millions of people everyday - Amazon - bring Firefox to a crawl and the devs instead of fixing the problem keep adding video chat to the bloated thing? It's just insane.
I only have this problem with Amazon. I often have a hundred tabs open on various sites, all of which runs and fits fine on my overclocked CPU and 16 GBytes of RAM. But if I open more than one or two Amazon tabs, even if that's all I have open, life slows to a crawl and my Firefox CPU usage goes to 101% of one CPU core. It has been this way for many months with Amazon, across multiple machines and after clean installations on brand new disks of Ubuntu, Debian and Windows.
Amazon is trying to do too much active updating on each page, or some such. The best recommendations I've seen from them are to clear my cache. Does a fresh install onto a new blank Samsung Pro 850 SSD count as clearing cache ?
Using other browsers might allow -other- tabs than Amazon to proceed with less blockage, if those other browsers have better multi-threading support, but the Amazon site remains slow if I have more than one Amazon tab open, whether on Firefox, Internet Explorer, Opera, or Chrome.
Earth to Amazon... we have a problem down here in Houston.
The MDHearingAid Acoustitone PRO Hearing Aid is $179.99 (at http://amzn.com/B00431MFHS). You can get a custom earpiece for another $75 from averysound.com.
You make a good point about absentee voting being
a tradeoff, between enabling the vote for those
who couldn't make it to a polling place, and
reducing the assurance that ballots are cast in
secret.
An idea... it seems that one important element of
a proper election is the secret ballot, and that this can best be done by having the ballot cast in
a controlled public and publicized place - so that
the public can see that ballots are cast in secret
without the opportunity for coercion.
But this doesn't mean that all ballots have to
be cast locally. How about something analogous
to a Public Notary, which is a low tech answer
to a vaguely similar problem.
In the case of elections, it would mean public
voting places, at which you could cast a ballot
for any election, anywhere. A separate and
unbiased (er eh... is such possible?) organization, perhaps one of several selected
competitively, would operate polling places,
at announced times and places, open to inspection
by all parties, ensuring ballots were cast in
secret, using electronic balloting to enable
the submission of a ballot for any election being
held that day.
This would solve (in a low tech, 'good enough' way)
the problem of ensuring a secret ballot.
The sweet lady who checks me into the local
school gym to vote could just as well be
doing that for someone casting a ballot in an
election clear across the country. The
various other mechanisms being discussed would
still be needed to provide the other desired
properties of such elections.
which was referenced by the first posting in
this thread. The first posting in this thread
simplified things too much, blaming insecure
internets and such. But the paper which it
referenced goes into some detail the various
alternatives and issues with computerized voting.
> but at that point how do we know if it's doing all the dirty bits correctly?
You don't, unless you have some expertise in
crypto software, and access to the exact source
code used for the software.
This is a serious problem with fancy computerized
voting schemes. One has to trust one's government
and its contractors to be technically astute and
beyond moral reproach.
One of the few things most Democrats and Republicans can agree on is that they don't
always have such trust.
> Some may be invalid, others are successfully
> forged. Unless the "unique number" is a
> randomly-generated 128+ bit number, forgery
> is quite possible.
Well... duh... make it 128 or 256 or whatever
bits then!
When the number of possible combinations
would require a trillion computers operating
at a trillion cycles per second a trillion years
to enumerate, then the risk that someone can find
a duplicate is less than the risk that my dog
will be elected President via write-in ballots.
And my dog died 40 years ago.
The above example could provide 2 to the power
of 145 compute cycles, if my math is right.
So 145 bits should be plenty.
> given a voter's unique number, generate
> all possible votes and match the results.
See above - use enough bits, and this will take
too long (as in way longer than our Sun
will be around).
> decrypt the vote
Just as with real, in-person, voting, the envelope
that identifies the voter has to be separated from
the ballot contents, once the voter is checked off
the rolls, before the ballot is tabulated.
In this case, the envelope is decoded and printed
out, showing the voter identity (equivalent to my
declaring my identity at the polling place), but
the ballot contents is included as a cryptic
number, undecipherable to the poll workers. These
are split, the ballot in one box, the identity
in the other, in front of workers from both parties (still a possibility of ballot stuffing
and spoilage,
just as in real life). Once split, the ballots
are sent on to be decoded and tabulated, without
any identifying trace left of who cast them.
> No paper trail... Actually there is no proof either way.
Bingo - nor should there be. The secret ballot
is required to protect against vote selling or
coercion.
> It is very difficult to do this electronically.
On this part you are correct. It would be especially difficult to do in a way that the
average poll worker (someone's grandmother,
perhaps) could easily audit, and such that the
below-average voter could understand and trust
as being honest and fair.
A system that only computer scientists with
a good mathematics background can understand
is not a workable system, regardless of how
solid the math behind it.
I am not saying it can really be done anytime
soon, in a way that ordinary folks can understand
and trust. Nor am I saying that our current
governmental beauracracies can do it in a way that
is technically competent - crypto is not your
average politicians strong suit.
My original point was that saying it can't be
done because the PC's and internet being used
are insecure is wrong. It still can't be done
right, just for different reasons.
However, regardless of whether a secure voting
mechanism can be implemented over the internet,
this one is in deep doo-doo:
I doubt that our institutions are capable of
providing a secure voting mechanism without much
trial and error (at the expense of our elections)
Those who make the most use of election fraud
now have much to gain from claiming that this
voting system is allowing the _other_ side to
steal elections. It puts the heat on the other
side, and tends to delay the acceptance of these
new systems until the dishonest figure out how
to hack them as well. Meanwhile, the same old
systems are used as before, with well known ways
of being stolen.
Elections are a public act, that depend on
the public trust. Most of the public can be
easily persuaded to distrust any electronic system that
can be imagined, especially one that is actually
subtle enough to be secure.
The article seems to be saying that because the
internet and the PC used for voting is insecure,
therefore the voting system must be insecure.
That part I don't agree with.
It is fundamentally possible to have secure communications over an insecure link. For example, each voter gets a unique number, encrypts their ballot using a common
public key inside a message encrypted using their
unique number. At election headquarters, votes
can be received by paper, email, or any other
insecure means of transmitting a thousand bytes
or so of data. Each received vote is printed out
with the outside portion decrypted to identify the
unique voter who sent it, so it can be checked
off against voter rolls, but the inside ballot
still as a cryptic number. A piece of verifiable
software can repeatedly reread the cryptic ballot
numbers off the pile of hardcopy ballots, to
produce repeatable election results, and the pile
of hardcopy ballots can be repeatedly checked against the voter rolls to ensure that each
ballot was cast by a valid voter, and each voter
voted at most once.
How do spammers get people's email addresses ?
http://www.faqs.org/faqs/net-abuse-faq/harvest/
It lists and describes the following mechanisms:
1. From posts to UseNet with your email address. 2. From mailing lists. 3. From web pages. 4. From various web and paper forms. 5. Via an Ident daemon. 6. From a web browser. 7. From IRC and chat rooms. 8. From finger daemons. 9. AOL profiles. 10. From domain contact points. 11. By guessing & cleaning. 12. From white & yellow pages. 13. By having access to the same computer. 14. From a previous owner of the email address 15. Using social engineering. 16. Buying lists from others. 17. By hacking into sites.
This FAQ also has a number of useful links to other spam resources.
This bug doesn't surprise me - the previous line
reads "EXCEPTION_RAISE...". One would expect
from the verb "RAISE" that it was going to jump
out of line right there, and that the next
line would be NOTREACHED.
A proper fix for this would change the name
of that EXCEPTION_RAISE macro to something
that doesn't suggest out of sequence execution.
Someone should grep through the source for lsh,
and see if there are any other places where after
this macro is called, the code really is expecting
execution to continue inline.
The campus restaurant is still open - though I
can't speak to its value - junk food for me.
I had promised myself to quit SGI when the Cokes
were no longer 25 cents. They're 50 cents as of
a year ago. I broke my promise... a family to
feed and the job market kinda sucks.
My handle comes from Python, the little sister to
Perl, and Cows, which were nice and warm in the
barn on a cold winters night in my youth.
Can't say as we were using the layoff this time to
get rid of deadwood. Everyone I know who got the
ax was quite lively and valued. We do have to run
a business, and that can be challenging at times.
Bishop's the best CEO I've seen in my 25 years;
he picked up a real challenge here, and is making
as well as could be hoped of it, with focus,
integrity, and good leadership.
And yes, we definitely layed off more than we
are hiring - I don't know the inside numbers,
but what I've seen reported looked like about
400 out of 4000 reduction in total employees.
Almost every company is hiring somewhere,
even the ones that are downsizing. Just because
you have to resize doesn't mean that there aren't
a couple of positions where you have openings.
Linux is doing well at SGI, your opportunity
might well be there.
I too am fond of TheBat!. Unlike many email programs you can have multiple things going on at once, replying, searching and reading. I run Win4Lin on top of Linux in good part to get TheBat!, having found no Linux email client that is as good (though Pine comes pretty close, for a curses style interface).
I wish the situation for USENET news clients was half as good. I've found nothing there, either Windows nor Linux, that I don't end up cursing frequently. Currently, I'm using trn, but it's a bitch.
Yes - I also use Astaro, on an old PC I setup as my home firewall. I've been delighted with it. It's easy to modify the firewall rules when I or my son have a new game that needs another hole punched in the wall.
Astaro seems to be one of the top downloads from SourceForge for firewalls.
Did you ever notice that you can tell a telemarketing call by the dead-air for a
few seconds after you pick up. I presume
they had some machine dialing, and only
after an apparent answer they connect the
next available person to the call.
So have a phone that
at first doesn't ring through, but picks up the
call quietly, and says, _quickly_, "Press One
to Continue".
If it is a person, they Press One, it rings through, and you or your answering machine answer.
If it is a telemarketer, they don't Press One because they are late to the prompt and end up talking to dead air until they get bored.
I too prefer Win4Lin now. My home machine has been running Win98 for a long time, because I needed Quicken (Q), and my wife needed Lotus Wordpro (LWP), or a similar polished word processor. I had tried converting to Linux, running VMWare for Q and LWP, but it always seemed more elaborate than I had time to master.
I had troubles getting Win4Lin installed, and like another person on this thread found the teech support for NeTraverse unable to help much. But its command line install scripts were simple enough to grok, and I muddled through.
I learned the following during the Win4Lin install to a Mandrake 8.0 running high security with a 433 MHz Celeron
and 256 Mb memory.
1) Don't try the install with a CDPATH lacking "." - the scripts contain cmds resembling "mkdir x; cd x" which fail if CDPATH is set, but lacking ".".
2) Don't try to install or run Win4Lin with your/var partition mounted "nodev", for Win4Lin needs numerous devices under/var/win4lin/mrgdev. Actually I made this path a symlink to another partition, just for Win4Lin, so I could continue to run/var mounted "nodev". I don't want to make it easy for hackers to put special device files in
/var/tmp.
3) If you're an old cheap skate like me, still installing Win98 as an upgrade from a set of Win3.1 floppies, then
allow 30 seconds between each floppy insertion when validating your Win 3.1 floppies to the Win98 installer. The
floppy change is missed, and there is a 30 second timeout buried in there somewhere.
Hopefully the above 3 tips will help someone.
I encouraged NeTraverse to add the above to their FAQ's (or actually fix code or docs), but given their recent layoffs and such, I don't expect much there.
Anyway, I am now happily running my home machine with Mandrake 8.0, still using Quicken, Lotus WordPro and TheBat mail client from Windows via Win4Lin, and InfoSelect running under Dosemu.
I tried converting my wife to Abiword, but the lack of anti-aliased font support, and a strange printing bug (hyphen char on line causes last char on line to print mangled) caused that effort to fail big time.
I have a Linux desktop icon for my wife that always switches to Win4Lin running full screen windows. The script behind the icon looks to see if Win4Lin is already running. If not, it invokes "fwin -auth &". If it is running,
it flips to the Virtual Terminal on which Win4Lin is presumably running. That is, it does a programmatic equivalent of Ctrl-Alt-F10, using essentially the following code:
#include <linux/vt.h>
main() {
int fd = open ("/dev/tty10", 2);
ioctl (fd, VT_ACTIVATE, 10);
}
If you're Win4Lin full screen is on VT 8 (typical Red Hat) instead of 10 (as in my Mandrake install) then replace
both '10' above with '8'.
Slashdot Mirror
Well, I'll be a monkey's uncle. The Amazon performance problem seems to be fixed. After months of muttering to myself about the problems with Amazon, it appears to be fixed. RIght now I have a half dozen Amazon tabs open in my Firefox browser (on a well equipped desktop PC with Debian) and performance is good. At the same time, I notice that when I hover my mouse over the "My Cart" icon, upper right of most pages, I no longer see a list magically appear of what I have in my shopping cart. I have to click on the Cart to go there. This may well be the sort of reasonable tradeoff that was required to get adequate performance - tune down the glitz a little.
This was exactly the reason that gave the final push to ditch Firefox for me as well. Seriously, how can a page that's seen by millions of people everyday - Amazon - bring Firefox to a crawl and the devs instead of fixing the problem keep adding video chat to the bloated thing? It's just insane.
I only have this problem with Amazon. I often have a hundred tabs open on various sites, all of which runs and fits fine on my overclocked CPU and 16 GBytes of RAM. But if I open more than one or two Amazon tabs, even if that's all I have open, life slows to a crawl and my Firefox CPU usage goes to 101% of one CPU core. It has been this way for many months with Amazon, across multiple machines and after clean installations on brand new disks of Ubuntu, Debian and Windows. Amazon is trying to do too much active updating on each page, or some such. The best recommendations I've seen from them are to clear my cache. Does a fresh install onto a new blank Samsung Pro 850 SSD count as clearing cache ? Using other browsers might allow -other- tabs than Amazon to proceed with less blockage, if those other browsers have better multi-threading support, but the Amazon site remains slow if I have more than one Amazon tab open, whether on Firefox, Internet Explorer, Opera, or Chrome. Earth to Amazon ... we have a problem down here in Houston.
The MDHearingAid Acoustitone PRO Hearing Aid is $179.99 (at http://amzn.com/B00431MFHS). You can get a custom earpiece for another $75 from averysound.com.
An idea ... it seems that one important element of
a proper election is the secret ballot, and that this can best be done by having the ballot cast in
a controlled public and publicized place - so that
the public can see that ballots are cast in secret
without the opportunity for coercion.
But this doesn't mean that all ballots have to be cast locally. How about something analogous to a Public Notary, which is a low tech answer to a vaguely similar problem.
In the case of elections, it would mean public voting places, at which you could cast a ballot for any election, anywhere. A separate and unbiased (er eh ... is such possible?) organization, perhaps one of several selected
competitively, would operate polling places,
at announced times and places, open to inspection
by all parties, ensuring ballots were cast in
secret, using electronic balloting to enable
the submission of a ballot for any election being
held that day.
This would solve (in a low tech, 'good enough' way) the problem of ensuring a secret ballot. The sweet lady who checks me into the local school gym to vote could just as well be doing that for someone casting a ballot in an election clear across the country. The various other mechanisms being discussed would still be needed to provide the other desired properties of such elections.
> a little bit more clearly.
That would take more time, space and effort than I can devote here. You might try reading the original document:
which was referenced by the first posting in this thread. The first posting in this thread simplified things too much, blaming insecure internets and such. But the paper which it referenced goes into some detail the various alternatives and issues with computerized voting.You don't, unless you have some expertise in crypto software, and access to the exact source code used for the software.
This is a serious problem with fancy computerized voting schemes. One has to trust one's government and its contractors to be technically astute and beyond moral reproach.
One of the few things most Democrats and Republicans can agree on is that they don't always have such trust.
I agree. My counter example suggests how insecure PC's and internet might be worked around, but it has other problems, such as you suggest.
However, absentee voting has this same problem!
> forged. Unless the "unique number" is a
> randomly-generated 128+ bit number, forgery
> is quite possible.
Well ... duh ... make it 128 or 256 or whatever
bits then!
When the number of possible combinations would require a trillion computers operating at a trillion cycles per second a trillion years to enumerate, then the risk that someone can find a duplicate is less than the risk that my dog will be elected President via write-in ballots.
And my dog died 40 years ago.
The above example could provide 2 to the power of 145 compute cycles, if my math is right. So 145 bits should be plenty.
> given a voter's unique number, generate
> all possible votes and match the results.
See above - use enough bits, and this will take too long (as in way longer than our Sun will be around).
> decrypt the vote
Just as with real, in-person, voting, the envelope that identifies the voter has to be separated from the ballot contents, once the voter is checked off the rolls, before the ballot is tabulated.
In this case, the envelope is decoded and printed out, showing the voter identity (equivalent to my declaring my identity at the polling place), but the ballot contents is included as a cryptic number, undecipherable to the poll workers. These are split, the ballot in one box, the identity in the other, in front of workers from both parties (still a possibility of ballot stuffing and spoilage, just as in real life). Once split, the ballots are sent on to be decoded and tabulated, without any identifying trace left of who cast them.
> No paper trail ... Actually there is no proof either way.
Bingo - nor should there be. The secret ballot is required to protect against vote selling or coercion.
> It is very difficult to do this electronically.
On this part you are correct. It would be especially difficult to do in a way that the average poll worker (someone's grandmother, perhaps) could easily audit, and such that the below-average voter could understand and trust as being honest and fair.
A system that only computer scientists with a good mathematics background can understand is not a workable system, regardless of how solid the math behind it.
I am not saying it can really be done anytime soon, in a way that ordinary folks can understand and trust. Nor am I saying that our current governmental beauracracies can do it in a way that is technically competent - crypto is not your average politicians strong suit.
My original point was that saying it can't be done because the PC's and internet being used are insecure is wrong. It still can't be done right, just for different reasons.
That part I don't agree with.
It is fundamentally possible to have secure communications over an insecure link. For example, each voter gets a unique number, encrypts their ballot using a common public key inside a message encrypted using their unique number. At election headquarters, votes can be received by paper, email, or any other insecure means of transmitting a thousand bytes or so of data. Each received vote is printed out with the outside portion decrypted to identify the unique voter who sent it, so it can be checked off against voter rolls, but the inside ballot still as a cryptic number. A piece of verifiable software can repeatedly reread the cryptic ballot numbers off the pile of hardcopy ballots, to produce repeatable election results, and the pile of hardcopy ballots can be repeatedly checked against the voter rolls to ensure that each ballot was cast by a valid voter, and each voter voted at most once.
There's an FAQ covering this:
How do spammers get people's email addresses ?
http://www.faqs.org/faqs/net-abuse-faq/harvest/
It lists and describes the following mechanisms:
1. From posts to UseNet with your email address.
2. From mailing lists.
3. From web pages.
4. From various web and paper forms.
5. Via an Ident daemon.
6. From a web browser.
7. From IRC and chat rooms.
8. From finger daemons.
9. AOL profiles.
10. From domain contact points.
11. By guessing & cleaning.
12. From white & yellow pages.
13. By having access to the same computer.
14. From a previous owner of the email address
15. Using social engineering.
16. Buying lists from others.
17. By hacking into sites.
This FAQ also has a number of useful links
to other spam resources.
A proper fix for this would change the name of that EXCEPTION_RAISE macro to something that doesn't suggest out of sequence execution.
Someone should grep through the source for lsh, and see if there are any other places where after this macro is called, the code really is expecting execution to continue inline.
I had promised myself to quit SGI when the Cokes were no longer 25 cents. They're 50 cents as of a year ago. I broke my promise ... a family to
feed and the job market kinda sucks.
My handle comes from Python, the little sister to Perl, and Cows, which were nice and warm in the barn on a cold winters night in my youth.
Can't say as we were using the layoff this time to get rid of deadwood. Everyone I know who got the ax was quite lively and valued. We do have to run a business, and that can be challenging at times.
Bishop's the best CEO I've seen in my 25 years; he picked up a real challenge here, and is making as well as could be hoped of it, with focus, integrity, and good leadership.
And yes, we definitely layed off more than we are hiring - I don't know the inside numbers, but what I've seen reported looked like about 400 out of 4000 reduction in total employees.
Linux is doing well at SGI, your opportunity might well be there.
Isaac R.........h you old dog you - pj here. ... better than you think.
SGI's still going
Take care, old buddy.
I too am fond of TheBat!. Unlike many email programs you can have multiple things going on at once, replying, searching and reading. I run Win4Lin on top of Linux in good part to get TheBat!, having found no Linux email client that is as good (though Pine comes pretty close, for a curses style interface).
I wish the situation for USENET news clients was half as good. I've found nothing there, either Windows nor Linux, that I don't end up cursing frequently. Currently, I'm using trn, but it's a bitch.
Yes - I also use Astaro, on an old PC I setup as
my home firewall. I've been delighted with it.
It's easy to modify the firewall rules when I
or my son have a new game that needs another hole
punched in the wall.
Astaro seems to be one of the top downloads from
SourceForge for firewalls.
A variation of this might work
Did you ever notice that you can tell a telemarketing call by the dead-air for a few seconds after you pick up. I presume they had some machine dialing, and only after an apparent answer they connect the next available person to the call.
So have a phone that at first doesn't ring through, but picks up the call quietly, and says, _quickly_, "Press One to Continue".
If it is a person, they Press One, it rings through, and you or your answering machine answer.
If it is a telemarketer, they don't Press One because they are late to the prompt and end up talking to dead air until they get bored.
I too prefer Win4Lin now. My home machine has been running Win98 for a long time, because I needed Quicken (Q), and my wife needed Lotus Wordpro (LWP), or a similar polished word processor. I had tried converting to Linux, running VMWare for Q and LWP, but it always seemed more elaborate than I had time to master.
/var partition mounted "nodev", for Win4Lin needs numerous devices under /var/win4lin/mrgdev. Actually I made this path a symlink to another partition, just for Win4Lin, so I could continue to run /var mounted "nodev". I don't want to make it easy for hackers to put special device files in
I had troubles getting Win4Lin installed, and like another person on this thread found the teech support for NeTraverse unable to help much. But its command line install scripts were simple enough to grok, and I muddled through.
I learned the following during the Win4Lin install to a Mandrake 8.0 running high security with a 433 MHz Celeron and 256 Mb memory.
1) Don't try the install with a CDPATH lacking "." - the scripts contain cmds resembling "mkdir x; cd x" which fail if CDPATH is set, but lacking ".".
2) Don't try to install or run Win4Lin with your
/var/tmp.
3) If you're an old cheap skate like me, still installing Win98 as an upgrade from a set of Win3.1 floppies, then
allow 30 seconds between each floppy insertion when validating your Win 3.1 floppies to the Win98 installer. The floppy change is missed, and there is a 30 second timeout buried in there somewhere.
Hopefully the above 3 tips will help someone.
I encouraged NeTraverse to add the above to their FAQ's (or actually fix code or docs), but given their recent layoffs and such, I don't expect much there.
Anyway, I am now happily running my home machine with Mandrake 8.0, still using Quicken, Lotus WordPro and TheBat mail client from Windows via Win4Lin, and InfoSelect running under Dosemu.
I tried converting my wife to Abiword, but the lack of anti-aliased font support, and a strange printing bug (hyphen char on line causes last char on line to print mangled) caused that effort to fail big time.
I have a Linux desktop icon for my wife that always switches to Win4Lin running full screen windows. The script behind the icon looks to see if Win4Lin is already running. If not, it invokes "fwin -auth &". If it is running,
it flips to the Virtual Terminal on which Win4Lin is presumably running. That is, it does a programmatic equivalent of Ctrl-Alt-F10, using essentially the following code:
#include <linux/vt.h>
main() {
int fd = open ("/dev/tty10", 2);
ioctl (fd, VT_ACTIVATE, 10);
}
If you're Win4Lin full screen is on VT 8 (typical Red Hat) instead of 10 (as in my Mandrake install) then replace
both '10' above with '8'.