As someone who studied under milner (undergrad; his classes were kind of dull), I do occasionally whip out the formal methods to prove things one. But by and large I write unit tests, because with junit you can automated regression testing; whereas for proofs, any change to the code and you need to reprove everything. Plus you have to make so many assumptions about the underlying system, & unit tests are great for testing those assumptions (like the XML parser works, the network is running, etc).
So what do I prove: that my threaded code is synchronized correctly. This is stuff that tests dont always catch, even on two-way or four-way systems, and I dont want to wait until ship time before finding I left a race condition or a deadlock.
Of course, once I am satisfied that the threading works & have to document it and say 'leave this alone'; becaus not enough developers can do proofs, it is a major maintenance risk.
I live in an old house (1840), in the UK, and when we had the floors up we found lead piping for gas lighting, which was the premium source of light before electicity came along.
After admiring the historical quaintess of century and a half old technology, we pulled it up and sold it for enough to cover some of the costs of the woodwork repairs, then laid down CAT5 (attenuation in stone is atrocious, especially for 802.11a, so CAt5 is the backbone).
I hope in another 150 years someone will find the cat5 wiring and find it equally quaint, as they laugh at 100mbit bandwidth and IPv4 net addresses. At least I hope so -as I doubt they will find as much resale value in the wires as we did in the lead pipes.
Fibre is worthless. Copper valuable enough on long haul connections that cable -theft is an issue in places like china. Fibre and cellphones are used there instead...
MAPI really means 'MAIL API'; its the API that apps wanting to talk to a mail program use.
Exchange is simply a MAPI implementation, using some secret RPC mechanism that is almost a reference 'how not to write a distributed application' design. I guess it is probably documented in the NDA-only, RAND-licensed settlement protocols, but that is no use whatsoever.
Maybe we could use OLE automation to talk to outlook, and bridge into the exchange database that way. That is probably what the WindowsCE synchonisation stuff does, after all.
the first claim is way too broad and vague; the later ones focus too far on voicemail systems.
And: the best bit, the sequence is too tight. It says
1. get id known to the approving entity (like a visa card) from the customer
2. get a debit value from the vendor
3. send (id,debit) to the auth system
4. get approval from the auth system
5. forward that to the vendor.
Here are some easy workarounds
a) get a range like (4.00 and 4.03) from the vendor; pick one.
b) get a limit from the auth system, then OK it with the vendor while sending the full debit to the auth system (bit of a race condition here)
c) get auth for a bit more than the debit value from the auth system
d) always grant provisional approval to the vendor, then send a refusal if it wasnt authed; 'silence' within a timeout implies granted. brittle.
e) dont forward auth information at all. Send the vendor an authorisation token (maybe a URL) that is handled by the auth system, and it is up to the vendor to probe that URL to complete the auth.
(e) is my favourite, as it gives the vendor the extra ability to make more SOAP/REST calls on the URL, like commit the transaction, or cancel it.
Overall then: easy enough to work around, no need to pony up big numbers of $$.
the patent doesnt start of with telecoms. Claim 1 says 'a communications routing system of a type which functions generally to establish connections with arbitrary ones of a plurality of entities'. That could cover yodelling credit card info over the alps as much as anything else.
Later on it gets into 'special phone numbers' -the core of the patent is really about humans-on-phones, but unless those primary claims are junked then we are in trouble.
Claim 5 says "the communications routing system is a switching portion of a telephone system;"; the internet may or may not be this, depending upon implementation details. If you disprove this and the previous claim 'any routing system' then ebay have a problem.
On the bright side, I think claim 1 is way too broad. There are lots of places -like many italian shops- where the payment person is split from the billing person, so the mechanism claimed for has existed in the physical world for ages.
qwest certainly provide some semblance of a local phone service here in Oregon. that is, if your expectation of a phone service is based on previous experience in a former communist state or region undergoing extended civil war.
What always irritates me is that their on-line billing system shuts down on weekend evenings. That is right, you cannot pay your bill from 8pm saturday. Now I know I should be going out and having fun on a saturday, but sometimes I dont, and on those days I'd like to pay my phone bill, rather than get a 'service unavailable message'.
I use Win2K in a VMWare box at work too -it has less advertising and overhead than XP after all.
What I am actually doing is a sort of device drive, and we have to support Admin-only install, as it is the only way to install something with the right rights to send commands down to a physical device.
but we are testing that untrusted users, including XP users can run the system once it is running, which does require a lot of hackery indeed. Oh, for a setuid bit on code.
XP home acts as you described, so, I think, does XP professional. Win2K3 server is the only version of windows that could be viewed as vaguely locked down.
A fundamental issue with windows is that everything usually needs admin rights to install. Look at these directX slides (esp slide 23) (sorry, PPT, but openoffice handles it), where they come out and say 'to install a game you need to be an admin'.
So you cannot even create a safe sandbox for a kid to do their thing, when they cannot plug in a game CD and expect it to run unless they have admin rights. What on earth were they thinking?
I think they thought they could break even and make money on the games. But sony are more ruthless.
PS/2 is a major slice of Sony's income -what 60%? They cannot afford to lose and are prepared to do whatever it takes.
MS are blessed with lots in the bank, and more coming in. Even then, a money sink in the home business unit can't be appreciated by the rest of the company. They are lucky MSN is just about breaking even these days.
point taken. 'If the HDD was useful then it could have been worth the money'. But as they didnt, it was $70 of cost that is hard to redeem.
Hard disks are special as their base price bottoms out above $50 -instead of their price falling, what you can get for your money increases. But while desktops and servers gain from the capacity increases, embedded systems just suffer.
You can take BSD code and GPL it on a branch, by retaining the BSD rules and adding GPL stuff.
Which is why JBoss bundles so much apaches stuff. They do seem to leave the Apache license on source they dont change; I dont know what they do with stuff they cut and paste in to their own modules. Though judging by the XLevel example above they may take the SCO route: forget the origin and then get the lawyers to start threating people about it.
yes. but that plan had one small flaw -it was bollocks.
Sony used custom Si with the same die area as Itanium1, yet could afford to pull it off by selling in the millions.
MS thought that by reusing PC kit they could get in the business easily (true), and ride the continual fall in PC part cost. Unfortunately, PC parts had had their cost already sucked out of them, apart from the effective 5% a month cost reduction of the Si parts. HDD and the DVD dont have much cost reduction at all, so that HDD is $70 of rotating iron whose cost is fixed. The best bit: Sony also rode the fall in Si parts, didnt have an HDD to provide fixed cost and can cut the selling price of the PS/2 whenever their spreadsheet hints that MS may be about to break even on hardware.
I think the biggest mistake of MS was thinking they could sell the hardware at a loss and make money on the games. The trick is to sell the hardware at a profit and make even more money on the games. Sony do that. Adding the HDD was another error. All it does is replicate DLL hell and add the Bill Of Materials of the box.
It certainly needs an XML (SAX) parser; I dont know what its exact dependencies are. Axis Java is primarily tested on Xerces, crimson and the oracle one is trouble.
One problem with SOAP is that it (especially doc/lit calls) depends very heavily on XML Schema rather than DTDs. As XSD is an overdesigned by committee nightmare, SOAP is very fussy about parsers.
when do sysadmins use messenger service anyway? I have never come across anything other than a printer saying 'add more paper'. We have this thing called 'email' for central messages.
Historically it exists to let you know that a printer down the hall is out of paper. So behind the firewall, it almost has a role.
But there are three flaws
One: there is no security or authentication, lending it spammable.
Two: because XP ships with no firewall, even XP home, it is wide open.
Three: whoever wrote it didnt understand about buffer overflows, rendering it a security hole.
It is a symptom of a problem: There is a lot of code in WinXP, a lot of it is 'corporate code' -assumes a LAN and trust- and not enough of it has been tested against malicious hackers.
If I wanted to find a new hack into a WinXP box, I'd follow a route that seems to work well -look at those embedded services that ship powered on. I'd go for the very old -written before security mattered- and for the very new -not tested yet. Oh, what the heck, lets go for any service. They just can't code them properly, can they?
that loss of vendor support is the biggest threat. RedHat linux had the advantage of being effectively the standard distro for add-on things, so binary apps -vmware, nvidia drivers and our sites ipsec stack to name the three I use- all come with official versions for the OS. I dont know if they will all support fedora themselves, or whether they will expect me to move to RH enterprise.
Same goes for hardware in general, 'red-hat linux' was the check box item that US hardware vendors aimed for, if they thought about linux at all. Still, maybe this will force them to think more broadly about what it means to support linux
That is a good point. They have already changed HTTP behaviour. If you write some hot new HTTP successor app, how long before they decide to answer failed lookups with their marketing front end, rather than valid data.
What if they started to reply to senders with suggestions for valid email addresses, maybe with adverts for ink cartridges at the bottom.
What if they cached all to and from addresses to add them to their list of 'consenting' users.
Verisigns perspective was if it is technically feasible, they are prepared to do it.
Ok. I submitted , I could have cached a PDF or SVG copy somewhere, but not my home server as it would have died. Apache.org maybe.
I use openoffice.org myself, so could handle the formats, and am reasonably immune from the problem.
Next time I submit I will identify PDF and SVG copies of the docs.
The transcript is in plain ascii, and very well transcribed. Its nice to see ICANN holding some meetings in public, along with a SpamAssassin rep joining in with some good comments.
Yes, it would be funny if it wasnt true. In exchange for the search revenue they are prepared to break everything.
Only one person in the transcript (read it, if you havent), asked 'what about the app developers -dont you have an implicit contract not to return wildcards', and Verisign replied "we only care about the standards", meaning no.
So the people who write the apps that make DNS lookups dont get consulted, dont get listened to, just get given extra work.
Yet if hadnt been for the app developers, the DNS business would be nothing. They should recognise that we dont need to use DNS, and if they try hard we can use alternate directory mechanisms: DOI, google, UDDI, IM directories even napster usernames are alternatives.
Maybe verisign are planning on doing the education. I can image 30 second TV ads where a third tier movie star explains that 'sometimes, "connection refused" means "unknown host". But I cannot image verisign paying for it.
Slashdot Mirror
As someone who studied under milner (undergrad; his classes were kind of dull), I do occasionally whip out the formal methods to prove things one. But by and large I write unit tests, because with junit you can automated regression testing; whereas for proofs, any change to the code and you need to reprove everything. Plus you have to make so many assumptions about the underlying system, & unit tests are great for testing those assumptions (like the XML parser works, the network is running, etc).
So what do I prove: that my threaded code is synchronized correctly. This is stuff that tests dont always catch, even on two-way or four-way systems, and I dont want to wait until ship time before finding I left a race condition or a deadlock.
Of course, once I am satisfied that the threading works & have to document it and say 'leave this alone'; becaus not enough developers can do proofs, it is a major maintenance risk.
I live in an old house (1840), in the UK, and when we had the floors up we found lead piping for gas lighting, which was the premium source of light before electicity came along.
After admiring the historical quaintess of century and a half old technology, we pulled it up and sold it for enough to cover some of the costs of the woodwork repairs, then laid down CAT5 (attenuation in stone is atrocious, especially for 802.11a, so CAt5 is the backbone).
I hope in another 150 years someone will find the cat5 wiring and find it equally quaint, as they laugh at 100mbit bandwidth and IPv4 net addresses. At least I hope so -as I doubt they will find as much resale value in the wires as we did in the lead pipes.
It depends on the cable.
Fibre is worthless. Copper valuable enough on long haul connections that cable -theft is an issue in places like china. Fibre and cellphones are used there instead...
MAPI really means 'MAIL API'; its the API that apps wanting to talk to a mail program use.
Exchange is simply a MAPI implementation, using some secret RPC mechanism that is almost a reference 'how not to write a distributed application' design. I guess it is probably documented in the NDA-only, RAND-licensed settlement protocols, but that is no use whatsoever.
Maybe we could use OLE automation to talk to outlook, and bridge into the exchange database that way. That is probably what the WindowsCE synchonisation stuff does, after all.
-steve
the first claim is way too broad and vague; the later ones focus too far on voicemail systems.
And: the best bit, the sequence is too tight. It says
1. get id known to the approving entity (like a visa card) from the customer
2. get a debit value from the vendor
3. send (id,debit) to the auth system
4. get approval from the auth system
5. forward that to the vendor.
Here are some easy workarounds
a) get a range like (4.00 and 4.03) from the vendor; pick one.
b) get a limit from the auth system, then OK it with the vendor while sending the full debit to the auth system (bit of a race condition here)
c) get auth for a bit more than the debit value from the auth system
d) always grant provisional approval to the vendor, then send a refusal if it wasnt authed; 'silence' within a timeout implies granted. brittle.
e) dont forward auth information at all. Send the vendor an authorisation token (maybe a URL) that is handled by the auth system, and it is up to the vendor to probe that URL to complete the auth.
(e) is my favourite, as it gives the vendor the extra ability to make more SOAP/REST calls on the URL, like commit the transaction, or cancel it.
Overall then: easy enough to work around, no need to pony up big numbers of $$.
the patent doesnt start of with telecoms.
Claim 1 says 'a communications routing system of a type which functions generally to establish connections with arbitrary ones of a plurality of entities'. That could cover yodelling credit card info over the alps as much as anything else.
Later on it gets into 'special phone numbers' -the core of the patent is really about humans-on-phones, but unless those primary claims are junked then we are in trouble.
Claim 5 says "the communications routing system is a switching portion of a telephone system;"; the internet may or may not be this, depending upon implementation details. If you disprove this and the previous claim 'any routing system' then ebay have a problem.
On the bright side, I think claim 1 is way too broad. There are lots of places -like many italian shops- where the payment person is split from the billing person, so the mechanism claimed for has existed in the physical world for ages.
qwest certainly provide some semblance of a local phone service here in Oregon. that is, if your expectation of a phone service is based on previous experience in a former communist state or region undergoing extended civil war.
What always irritates me is that their on-line billing system shuts down on weekend evenings. That is right, you cannot pay your bill from 8pm saturday. Now I know I should be going out and having fun on a saturday, but sometimes I dont, and on those days I'd like to pay my phone bill, rather than get a 'service unavailable message'.
I use Win2K in a VMWare box at work too -it has less advertising and overhead than XP after all.
What I am actually doing is a sort of device drive, and we have to support Admin-only install, as it is the only way to install something with the right rights to send commands down to a physical device.
but we are testing that untrusted users, including XP users can run the system once it is running, which does require a lot of hackery indeed. Oh, for a setuid bit on code.
XP home acts as you described, so, I think, does XP professional. Win2K3 server is the only version of windows that could be viewed as vaguely locked down.
A fundamental issue with windows is that everything usually needs admin rights to install. Look at these
directX slides (esp slide 23) (sorry, PPT, but openoffice handles it), where they come out and say 'to install a game you need to be an admin'.
So you cannot even create a safe sandbox for a kid to do their thing, when they cannot plug in a game CD and expect it to run unless they have admin rights. What on earth were they thinking?
I think they thought they could break even and make money on the games. But sony are more ruthless.
PS/2 is a major slice of Sony's income -what 60%? They cannot afford to lose and are prepared to do whatever it takes.
MS are blessed with lots in the bank, and more coming in. Even then, a money sink in the home business unit can't be appreciated by the rest of the company. They are lucky MSN is just about breaking even these days.
point taken. 'If the HDD was useful then it could have been worth the money'. But as they didnt, it was $70 of cost that is hard to redeem.
Hard disks are special as their base price bottoms out above $50 -instead of their price falling, what you can get for your money increases. But while desktops and servers gain from the capacity increases, embedded systems just suffer.
You can take BSD code and GPL it on a branch, by retaining the BSD rules and adding GPL stuff.
Which is why JBoss bundles so much apaches stuff. They do seem to leave the Apache license on source they dont change; I dont know what they do with stuff they cut and paste in to their own modules. Though judging by the XLevel example above they may take the SCO route: forget the origin and then get the lawyers to start threating people about it.
yes, the UK ad standards people are far fussier about advertising, and it makes a difference in what you get
-car ads cant glorify speed or dangerous driving (you can only show legal stuff)
-politicians arent allowed to advertise on TV at all. ('cos they all lie?)
-you need to be able to substantiate your claims.
-RSA inc once got chastised for saying their crypo solutions were better than OSS versions.
Whereas in the US, nike claims that freedom of speech laws allow it to lie about how its subcontracted workers are treated. Quelle difference.
yes. but that plan had one small flaw -it was bollocks.
Sony used custom Si with the same die area as Itanium1, yet could afford to pull it off by selling in the millions.
MS thought that by reusing PC kit they could get in the business easily (true), and ride the continual fall in PC part cost. Unfortunately, PC parts had had their cost already sucked out of them, apart from the effective 5% a month cost reduction of the Si parts. HDD and the DVD dont have much cost reduction at all, so that HDD is $70 of rotating iron whose cost is fixed. The best bit: Sony also rode the fall in Si parts, didnt have an HDD to provide fixed cost and can cut the selling price of the PS/2 whenever their spreadsheet hints that MS may be about to break even on hardware.
I think the biggest mistake of MS was thinking they could sell the hardware at a loss and make money on the games. The trick is to sell the hardware at a profit and make even more money on the games. Sony do that. Adding the HDD was another error. All it does is replicate DLL hell and add the Bill Of Materials of the box.
Interesting. Maybe this is why hp offered that indemnity thing.
'J2EE' is a very loaded term. I would not point you near Enterprise Java Beans as despite the hype, they make little sense.
Java Servlets are simple and easy to work with; you can get started with them and JSP pages (or Velocity) pretty simply.
bite the bullet, go on, learn a bit of Java. It is easier than learning C++.
It certainly needs an XML (SAX) parser; I dont know what its exact dependencies are. Axis Java is primarily tested on Xerces, crimson and the oracle one is trouble.
One problem with SOAP is that it (especially doc/lit calls) depends very heavily on XML Schema rather than DTDs. As XSD is an overdesigned by committee nightmare, SOAP is very fussy about parsers.
when do sysadmins use messenger service anyway? I have never come across anything other than a printer saying 'add more paper'. We have this thing called 'email' for central messages.
Historically it exists to let you know that a printer down the hall is out of paper. So behind the firewall, it almost has a role.
But there are three flaws
One: there is no security or authentication, lending it spammable.
Two: because XP ships with no firewall, even XP home, it is wide open.
Three: whoever wrote it didnt understand about buffer overflows, rendering it a security hole.
It is a symptom of a problem: There is a lot of code in WinXP, a lot of it is 'corporate code' -assumes a LAN and trust- and not enough of it has been tested against malicious hackers.
If I wanted to find a new hack into a WinXP box, I'd follow a route that seems to work well -look at those embedded services that ship powered on. I'd go for the very old -written before security mattered- and for the very new -not tested yet. Oh, what the heck, lets go for any service. They just can't code them properly, can they?
yes,
that loss of vendor support is the biggest threat. RedHat linux had the advantage of being effectively the standard distro for add-on things, so binary apps -vmware, nvidia drivers and our sites ipsec stack to name the three I use- all come with official versions for the OS. I dont know if they will all support fedora themselves, or whether they will expect me to move to RH enterprise.
Same goes for hardware in general, 'red-hat linux' was the check box item that US hardware vendors aimed for, if they thought about linux at all. Still, maybe this will force them to think more broadly about what it means to support linux
That is a good point. They have already changed HTTP behaviour. If you write some hot new HTTP successor app, how long before they decide to answer failed lookups with their marketing front end, rather than valid data.
What if they started to reply to senders with suggestions for valid email addresses, maybe with adverts for ink cartridges at the bottom.
What if they cached all to and from addresses to add them to their list of 'consenting' users.
Verisigns perspective was if it is technically feasible, they are prepared to do it.
Ok. I submitted , I could have cached a PDF or SVG copy somewhere, but not my home server as it would have died. Apache.org maybe.
I use openoffice.org myself, so could handle the formats, and am reasonably immune from the problem.
Next time I submit I will identify PDF and SVG copies of the docs.
The transcript is in plain ascii, and very well transcribed. Its nice to see ICANN holding some meetings in public, along with a SpamAssassin rep joining in with some good comments.
Yes, it would be funny if it wasnt true. In exchange for the search revenue they are prepared to break everything.
Only one person in the transcript (read it, if you havent), asked 'what about the app developers -dont you have an implicit contract not to return wildcards', and Verisign replied "we only care about the standards", meaning no.
So the people who write the apps that make DNS lookups dont get consulted, dont get listened to, just get given extra work.
Yet if hadnt been for the app developers, the DNS business would be nothing. They should recognise that we dont need to use DNS, and if they try hard we can use alternate directory mechanisms: DOI, google, UDDI, IM directories even napster usernames are alternatives.
yeah, it upset me no end too.
Maybe verisign are planning on doing the education. I can image 30 second TV ads where a third tier movie star explains that 'sometimes, "connection refused" means "unknown host". But I cannot image verisign paying for it.
The problem lexmark had here was refillers patching the (somewhat convoluted) Si that came with the cartridges to say 'this is full again'.
They probably couldnt use patent or copyright on the cartridge themselves as they were their own cartridges.
I dont know anyone else who has tried to make refilling illegal, which is what this action was an attempt to do.
Next question: did Dell put them up to it? Dell rebadge Lexmark printers, and only sell ink online, so may have more to lose...