start treating it like the need-to-regulate-speed resource like highways?
ROFLMAO we restrict speeds on highways for safetey, fuel efficiency and/or to generate revenue from speeding fines, not for any reason to do with capacity.
Oh and while we freqently reffer to bandwidth as "speed" it is totally different from the speed of a vehicle (that is more comparable to latency).
they need to start offering more realistic packages for light, medium and heavy users or something. Say about 1.5MB for light or email only users, 5MB or so for medium users and gamers, and 15MB for high users, and the just go up from there for ultra heavy users or biz class users who need those kinds of packages. I guarantee there will be people who will be perfectly happy paying $100 a month for 25MB
There may be some but mostly I'd rather have a faster connection with a (reasonable) cap so I can download what I want as quickly as possible. Most people (hoarder pirates excepted) don't download anywhere close to 24/7.
I'm sure for low end stuff they are fine but they really aren't suitable for a project like this IMO. They don't offer high enough layer counts or fine enough design rules.
A couple of other niggles I notice that aren't fatal but don't exactly inspire confidence.
* their specs fail to list one of the most critical design rules, the minimum annular ring. * they do not make it clear whether they will actually accept designs that use the (marginally smaller) metric sizes in brackets. This matters a lot because modern components and designs are metric.
Granted, I do not know much about PCB classifications
Obviously
I would think DIY open source hardware would not be designed so to an extremely narrow physical layout which could not be done by hand.
Sure, the problem is there are very few* parts that are both competitive functionally and available in packages that don't require very fine PCBs. Even if there are who is realistically going to want a smartphone that is the size of an old brick phone.
IIRC on the raspberry Pi they were trying to route it on four layers but had to settle for six and they still couldn't bring out most of the IO and that was on a board that had quite a bit of clear space to fan things out.
Start paying for more layers, fine design rules, blind vias etc and it would not at all surprise me to have a prototype board cost a couple of hundred dollars for the PCB and if multiple BGAs are involed for it to cost the same again for assembly.
* Apparently allwinner have recently produced one.
Yes it will take a while to send stuff by ship from china but small packages can be flown over in a few days. If it's taking weeks then either your supplier is using it as an excuse or you were too cheap to pay for fedex.
The bigger problem is there seems to be a culture of "do as little as you can get away with" rather than "do what the customer specified" and from what I can gather you really need someone who works for you (not for your contract manufacturer) and is fluent in both languages and comfortable with both cultures on the ground in the production environment to counter that.
The US government made a strategic stockpile of helium (driving up the price at the time) when they thought they would need it for airships. Now that airships are no longer considered of military interest are selling off that stockpile depressing the price.
The bit you are missing is that the reason He prices are so low is because the US government has been selling off their stockpile (the US government stockpiled helium when they thought airships would be strategically important).
I don't think the government should try to impose a price floor on all helium sales but I think it would be perfectly reasonable for them to put a price floor on selling off their own reserves.
One of the big problems with DSL and to a lesser extent cable connections is they often have lousy upstream. Fine if all you do is consume content but painful when you produce something big and want to upload it.
Also while small files may download "instantly" at 15 mbit/s large files certainly don't. Grabbing a new version of an install CD/DVD still takes noticable time.
Having said all that I don't do any of theese things often enough to justify the higher cost of a faster connection.
GSM networks went through the exact same hardware problem 10 years ago when they transitioned from TDMA-based GSM/GPRS/EDGE to WCDMA-based UMTS/HS(D|U)PA(+).
Hmm
My understanding was that UMTS has mostly the same protocol stack as GSM (with a new physical layer) so from the networks point of view there isn't much difference between a phone moving between 2G and 3G than there is with a phone moving between cells. Is that understanding wrong?
You are wrong, for non-commercial use the GPL allows to just point to an upstream source if you used it unmodified.
Not in general it doesn't, there are a few special cases but they don't seem to apply here.
If you received a written offer of source then in some cases (for GPL2 it merely had to be noncommercial, for GPLv3 it has to be both noncommerical and occasional) you can supply a copy of that written offer of source but if that written offer never existed in the first place (which unless you received binaries on a physical medium it probablly didn't) then this isn't an option for you.
GPLv3 gives you the option of pointing the user to a different place to download the source. However there are a number of strings attached. Firstly you have to provide "clear directions" to where the source can be found. Secondly it is your responsibility to ensure that said source remains available which means you are putting yourself at leagal risk if you rely on a server you have no contractual relationship with and hence could dissappear without notice. Furthermore GPLv2 DOES NOT give you this option, it specifically states that source and binary must be distributed from the "same place" and linux is still GPLv2 only.
IMO minified JS is not much different in this regard from java or flash. Both Java and Flash can be readilly decompiled back to something resembling source code.
fc00::/8 should not be used until and unless the internet standards processes specify how it is to be used. IIRC there have been serveral proposals for allocation but none have yet been standardised.
Also, have site unique addresses - the fc00:/8 - been deprecated, or can one use them if one can come up w/ a mechanism of guaranteeing global uniqueness?
The intention with fc00::/8 was that they would be gauranteed unique (through a registration body) but not routable on the public internet. However no such registration body has yet been set up so any use of fc00::/8 at the moment would be a violation of the addressing standards.
So for now we have to use fd00::/8 which is only "probabalisitically unique".
So unique local addresses - the fd00:/8 - are the ones to use in case of VPNs?
A VPN is no different from a physical connection.
If your VPN operates on the ethernet level bridging users to a LAN then it would be no different from them physically plugging into said LAN . So you can use link local, unique local or internet addresses as appropriate.
If your VPN operates on the IP level then you would have to use unique local addresses or internet addresses.
Each vlan is a seperate "link" from the perspective of IP so if something needs to be routed between vlans unique-local or internet addresses should be used.
Due to a combination of nimbyism and practical considerations it's very common to have multiple reactors at one site.
TMI-2 was where the famous accident happens and was shut down permanently due to massive internal radioactive contamination. Afaict TMI-1 has had minor incidents over the years but nothing that would require a permanent shutdown.
Afaict with USB mass storage on the nexus 7 you can read the block device without rooting but if you want to mount it properly either so you can use the content directly without copying it first or because you want to copy stuff back to the USB device you need to be rooted.
But if someone is transitioning to IPv6 and has so far been having a network, say 10.1.x.x in his office, what does he use?
There are really three choices. Internet addresses, unique local addreses and link local addresses. Lets consider them individually
If it is highly unlikely you will ever change IPs (e.g. you have provider independent space or are strongly tied to your provider such that it is unlikely you will ever need to renumber) you may as well use internet addresses for everything. That way you are gauranteed uniqueness and if you decide to make something accessible on the internet later it's just a matter of changing firewall rules. If you are a small buisness and are running IPv6 It is likely most of your machines will have an internet IPv6 IP but I would advise against using it for internal stuff to reduce the pain if/when your ISP changes your prefix.
Link local addresses are for stuff that you are pretty sure will never need to leave a single "link" (Unless it's split into VLANS an Ethernet network is a single "link" in this context regardless of how big it is).
That leaves unique local addresses, theese are what you should use when you need private addresses that can be routed internally within your own network and to other cooperating networks but not on the general internet.
Ick -- WSAAsynGetHostByName? In this day and age, you have a window handle lying around?
Old habbits die hard and all that but even if i'm not using it in new code there is still a need to adapt old code. So far the only way i'd found to do an IPv6 DNS lookup in the background of an event driven program using the windows DNS code is to create a thread to do it and have that thread notify the main thread when the lookup completes.
Have you looked at GetAddrInfoEx? In Windows 8/Server 2012, the DNS team added some Async features into it.
No I hadn't heard of it but there is no way i'm making my code dependent on win8 in the forseeable future.
as if Jimmy Wales has any real authority on Wikipedia any more
Right now the foundation board has 10 memebers 4 of which were apointed by the board, 5 by the community and one of which is wales himself.
So the real question is the loyalty of those 4 appointed board members. If they are more loyal to wales than the community it would be virtually impossible for the community to overrule him the most they could do is produce a tie in a board vote (interestingly the bylaws of the foundation don't seem to specify what happens in that case). OTOH if the appointed board memembers are more loyal to the community than to wales then the community could easilly override him.
Unless all systems attached are on the same subnet... And that plays hell with routing, causes congestion... There are reasons the 10.x is non-routed. It was aimed at large local networks - like a node cluster. Sucks when you have to go past a router. That requires routable numbers.
BS you can route subnets of 10.x on your private networks just fine. You just can't advertise them on the public internet.
The real problem comes when you are trying to link together a load of sites that are already using some part (or even all, it's a class A block so the default netmask is 255.0.0.0) of 10.0.0.0/8 for their local private network. It is likely that some users will need access to both the national network and existing local private networks. So if you use private IPs for your network you are stuck either trying to find a subset of 10.x that none of the sites are using (can work but there is no gaurantee there will be any such space and it's a problem if you want to add more sites later). Renumbering machines unrelated to your network at various sites so they don't clash with your network or using some horrible NAT hacks.
If you want a free v6 tunnel there are less elitist providers than sixxs. gogo6 (aka freenet6) even offer unauthenticated tunnels for individual machines* so you can just install their software and go.
Still I consider such tunnels as a tool for those who are interested in developing/testing IPv6 and maybe as a stopgap measure for a subset of end users who really need to reach v6 servers. If you are serious about v6 then you should be using a v6 capable ISP.
*If you want a prefix you have to create an account and authenticate to it but afaict creating an account with them is no big deal.
1: routers need to both understand IPv6 AND be able to forward it quickly. If the hardware forwarding engines can't handle the larger v6 addreses then a software update won't help you much. 2: any application software that needs to communicate over IPv6 needs to use the new v6 capable APIs. Converting software can be a pain either because it requires significant changes to support IPv6* or because the vendor is being a PITA and wants to tie in v6 support to an expensive upgrade you don't want. Or worse a v6 upgrade may simply not be available at all requiring the software to be replaced completely. 3: while windows XP has some IPv6 support it's not ready for an IPv6 only world.
*Some examples: * There is no direct IPv6 equivilent to WSAAsyncGetHostByName so any app that needs to perform lookups in the background will need to be converted to use threads for name lookups. * In windows XP it is not possible for one socket to listen for both IPv4 and IPv6 so apps that previously only listened on a single socket may well need design changes to allow them to listen on multiple sockets. * Any app that stores IPv4 addresses in a binary form or a fixed-width text feild will need data format changes
Windows support is nonexistent, M$ directed me to the PC maker upon contact -- and the PC maker support won't talk about nothing but the original configuration, WITHOUT service packs... (this was a couple of years ago, whne XP was better "supported").
Yeah, that is the downside of using OEM windows, product support is handled by the manufacturer of the PC who tend to suck at it. Afaict MS do offer real support but you will have to pay for it (either as part of a retail copy of windows, on a pay per incident basis or as part of a support contract)
Slashdot Mirror
Probablly cheaper just to keep the data in a ram array.
start treating it like the need-to-regulate-speed resource like highways?
ROFLMAO we restrict speeds on highways for safetey, fuel efficiency and/or to generate revenue from speeding fines, not for any reason to do with capacity.
Oh and while we freqently reffer to bandwidth as "speed" it is totally different from the speed of a vehicle (that is more comparable to latency).
they need to start offering more realistic packages for light, medium and heavy users or something. Say about 1.5MB for light or email only users, 5MB or so for medium users and gamers, and 15MB for high users, and the just go up from there for ultra heavy users or biz class users who need those kinds of packages. I guarantee there will be people who will be perfectly happy paying $100 a month for 25MB
There may be some but mostly I'd rather have a faster connection with a (reasonable) cap so I can download what I want as quickly as possible. Most people (hoarder pirates excepted) don't download anywhere close to 24/7.
Afaict you can get any speed you want pretty much anywhere you want if you are prepared to pay enough.
OOI Is that price for a "boardband" connection or for a serious connection with dedicated bandwidth, service level agreement etc?
I recently heard of BatchPCB
I'm sure for low end stuff they are fine but they really aren't suitable for a project like this IMO. They don't offer high enough layer counts or fine enough design rules.
A couple of other niggles I notice that aren't fatal but don't exactly inspire confidence.
* their specs fail to list one of the most critical design rules, the minimum annular ring.
* they do not make it clear whether they will actually accept designs that use the (marginally smaller) metric sizes in brackets. This matters a lot because modern components and designs are metric.
Granted, I do not know much about PCB classifications
Obviously
I would think DIY open source hardware would not be designed so to an extremely narrow physical layout which could not be done by hand.
Sure, the problem is there are very few* parts that are both competitive functionally and available in packages that don't require very fine PCBs. Even if there are who is realistically going to want a smartphone that is the size of an old brick phone.
IIRC on the raspberry Pi they were trying to route it on four layers but had to settle for six and they still couldn't bring out most of the IO and that was on a board that had quite a bit of clear space to fan things out.
Start paying for more layers, fine design rules, blind vias etc and it would not at all surprise me to have a prototype board cost a couple of hundred dollars for the PCB and if multiple BGAs are involed for it to cost the same again for assembly.
* Apparently allwinner have recently produced one.
Yes it will take a while to send stuff by ship from china but small packages can be flown over in a few days. If it's taking weeks then either your supplier is using it as an excuse or you were too cheap to pay for fedex.
The bigger problem is there seems to be a culture of "do as little as you can get away with" rather than "do what the customer specified" and from what I can gather you really need someone who works for you (not for your contract manufacturer) and is fluent in both languages and comfortable with both cultures on the ground in the production environment to counter that.
The US government made a strategic stockpile of helium (driving up the price at the time) when they thought they would need it for airships. Now that airships are no longer considered of military interest are selling off that stockpile depressing the price.
The bit you are missing is that the reason He prices are so low is because the US government has been selling off their stockpile (the US government stockpiled helium when they thought airships would be strategically important).
I don't think the government should try to impose a price floor on all helium sales but I think it would be perfectly reasonable for them to put a price floor on selling off their own reserves.
One of the big problems with DSL and to a lesser extent cable connections is they often have lousy upstream. Fine if all you do is consume content but painful when you produce something big and want to upload it.
Also while small files may download "instantly" at 15 mbit/s large files certainly don't. Grabbing a new version of an install CD/DVD still takes noticable time.
Having said all that I don't do any of theese things often enough to justify the higher cost of a faster connection.
GSM networks went through the exact same hardware problem 10 years ago when they transitioned from TDMA-based GSM/GPRS/EDGE to WCDMA-based UMTS/HS(D|U)PA(+).
Hmm
My understanding was that UMTS has mostly the same protocol stack as GSM (with a new physical layer) so from the networks point of view there isn't much difference between a phone moving between 2G and 3G than there is with a phone moving between cells. Is that understanding wrong?
You are wrong, for non-commercial use the GPL allows to just point to an upstream source if you used it unmodified.
Not in general it doesn't, there are a few special cases but they don't seem to apply here.
If you received a written offer of source then in some cases (for GPL2 it merely had to be noncommercial, for GPLv3 it has to be both noncommerical and occasional) you can supply a copy of that written offer of source but if that written offer never existed in the first place (which unless you received binaries on a physical medium it probablly didn't) then this isn't an option for you.
GPLv3 gives you the option of pointing the user to a different place to download the source. However there are a number of strings attached. Firstly you have to provide "clear directions" to where the source can be found. Secondly it is your responsibility to ensure that said source remains available which means you are putting yourself at leagal risk if you rely on a server you have no contractual relationship with and hence could dissappear without notice. Furthermore GPLv2 DOES NOT give you this option, it specifically states that source and binary must be distributed from the "same place" and linux is still GPLv2 only.
IMO minified JS is not much different in this regard from java or flash. Both Java and Flash can be readilly decompiled back to something resembling source code.
fc00::/8 should not be used until and unless the internet standards processes specify how it is to be used. IIRC there have been serveral proposals for allocation but none have yet been standardised.
Right
Also, have site unique addresses - the fc00:/8 - been deprecated, or can one use them if one can come up w/ a mechanism of guaranteeing global uniqueness?
The intention with fc00::/8 was that they would be gauranteed unique (through a registration body) but not routable on the public internet. However no such registration body has yet been set up so any use of fc00::/8 at the moment would be a violation of the addressing standards.
So for now we have to use fd00::/8 which is only "probabalisitically unique".
So unique local addresses - the fd00:/8 - are the ones to use in case of VPNs?
A VPN is no different from a physical connection.
If your VPN operates on the ethernet level bridging users to a LAN then it would be no different from them physically plugging into said LAN . So you can use link local, unique local or internet addresses as appropriate.
If your VPN operates on the IP level then you would have to use unique local addresses or internet addresses.
Each vlan is a seperate "link" from the perspective of IP so if something needs to be routed between vlans unique-local or internet addresses should be used.
Due to a combination of nimbyism and practical considerations it's very common to have multiple reactors at one site.
TMI-2 was where the famous accident happens and was shut down permanently due to massive internal radioactive contamination. Afaict TMI-1 has had minor incidents over the years but nothing that would require a permanent shutdown.
Afaict with USB mass storage on the nexus 7 you can read the block device without rooting but if you want to mount it properly either so you can use the content directly without copying it first or because you want to copy stuff back to the USB device you need to be rooted.
Read the documentation more carefully. The function itself was introduced in vista but it's functionality was expanded in win8.
Not that I care too much, I don't have any plans to stop supporting XP in my code in the foreseeable future.
But if someone is transitioning to IPv6 and has so far been having a network, say 10.1.x.x in his office, what does he use?
There are really three choices. Internet addresses, unique local addreses and link local addresses. Lets consider them individually
If it is highly unlikely you will ever change IPs (e.g. you have provider independent space or are strongly tied to your provider such that it is unlikely you will ever need to renumber) you may as well use internet addresses for everything. That way you are gauranteed uniqueness and if you decide to make something accessible on the internet later it's just a matter of changing firewall rules. If you are a small buisness and are running IPv6 It is likely most of your machines will have an internet IPv6 IP but I would advise against using it for internal stuff to reduce the pain if/when your ISP changes your prefix.
Link local addresses are for stuff that you are pretty sure will never need to leave a single "link" (Unless it's split into VLANS an Ethernet network is a single "link" in this context regardless of how big it is).
That leaves unique local addresses, theese are what you should use when you need private addresses that can be routed internally within your own network and to other cooperating networks but not on the general internet.
Ick -- WSAAsynGetHostByName? In this day and age, you have a window handle lying around?
Old habbits die hard and all that but even if i'm not using it in new code there is still a need to adapt old code. So far the only way i'd found to do an IPv6 DNS lookup in the background of an event driven program using the windows DNS code is to create a thread to do it and have that thread notify the main thread when the lookup completes.
Have you looked at GetAddrInfoEx? In Windows 8/Server 2012, the DNS team added some Async features into it.
No I hadn't heard of it but there is no way i'm making my code dependent on win8 in the forseeable future.
as if Jimmy Wales has any real authority on Wikipedia any more
Right now the foundation board has 10 memebers 4 of which were apointed by the board, 5 by the community and one of which is wales himself.
So the real question is the loyalty of those 4 appointed board members. If they are more loyal to wales than the community it would be virtually impossible for the community to overrule him the most they could do is produce a tie in a board vote (interestingly the bylaws of the foundation don't seem to specify what happens in that case). OTOH if the appointed board memembers are more loyal to the community than to wales then the community could easilly override him.
Unless all systems attached are on the same subnet... And that plays hell with routing, causes congestion... There are reasons the 10.x is non-routed. It was aimed at large local networks - like a node cluster. Sucks when you have to go past a router. That requires routable numbers.
BS you can route subnets of 10.x on your private networks just fine. You just can't advertise them on the public internet.
The real problem comes when you are trying to link together a load of sites that are already using some part (or even all, it's a class A block so the default netmask is 255.0.0.0) of 10.0.0.0/8 for their local private network. It is likely that some users will need access to both the national network and existing local private networks. So if you use private IPs for your network you are stuck either trying to find a subset of 10.x that none of the sites are using (can work but there is no gaurantee there will be any such space and it's a problem if you want to add more sites later). Renumbering machines unrelated to your network at various sites so they don't clash with your network or using some horrible NAT hacks.
If you want a free v6 tunnel there are less elitist providers than sixxs. gogo6 (aka freenet6) even offer unauthenticated tunnels for individual machines* so you can just install their software and go.
Still I consider such tunnels as a tool for those who are interested in developing/testing IPv6 and maybe as a stopgap measure for a subset of end users who really need to reach v6 servers. If you are serious about v6 then you should be using a v6 capable ISP.
*If you want a prefix you have to create an account and authenticate to it but afaict creating an account with them is no big deal.
A few places
1: routers need to both understand IPv6 AND be able to forward it quickly. If the hardware forwarding engines can't handle the larger v6 addreses then a software update won't help you much.
2: any application software that needs to communicate over IPv6 needs to use the new v6 capable APIs. Converting software can be a pain either because it requires significant changes to support IPv6* or because the vendor is being a PITA and wants to tie in v6 support to an expensive upgrade you don't want. Or worse a v6 upgrade may simply not be available at all requiring the software to be replaced completely.
3: while windows XP has some IPv6 support it's not ready for an IPv6 only world.
*Some examples:
* There is no direct IPv6 equivilent to WSAAsyncGetHostByName so any app that needs to perform lookups in the background will need to be converted to use threads for name lookups.
* In windows XP it is not possible for one socket to listen for both IPv4 and IPv6 so apps that previously only listened on a single socket may well need design changes to allow them to listen on multiple sockets.
* Any app that stores IPv4 addresses in a binary form or a fixed-width text feild will need data format changes
Windows support is nonexistent, M$ directed me to the PC maker upon contact -- and the PC maker support won't talk about nothing but the original configuration, WITHOUT service packs... (this was a couple of years ago, whne XP was better "supported").
Yeah, that is the downside of using OEM windows, product support is handled by the manufacturer of the PC who tend to suck at it. Afaict MS do offer real support but you will have to pay for it (either as part of a retail copy of windows, on a pay per incident basis or as part of a support contract)