Whether I would tie Lua to Apache, or Lighttpd, or nginx, or libhttp, or whatever, depends on other things. If it's a web site running other stuff, something like Apache might be the way to go. If I can run the Lua stuff on its own IP address or port, I could just run Lua the way it was designed, as an extension language for something else, like a lean HTTP daemon in C.
For now I have dedicated and colocated servers. Each "instance" is a whole machine with full root access. In the case of the colocated ones I own them, and I have VNC/IP access to the KVM on them, with a flash drive plugged in, and can recover or reload easily.
If I can't get root access, then I can't run my custom app.
But the configuration and operation of GRUB is a total bitch. I could not even find a document to describe the config file. And NO... I do NOT configure things by running programs. I have more involved setups that just running programs cannot figure out. For example MY installer scripts need to generate the config files, NOT run some program that can't run in that minimalist installer environment.
Syslinux fits like a glove. GRUB is like trying to wrap a coat around your hand to keep it warm.
As long as they don't have a usage cap facility that YOU can set your own limits on, in either computer based technical units of usage, or by dollars (even zero for the free tier), and in time frames other than just a whole month, then yeah... your money could end up disappearing into thin air.
I should be able to set usage restrictions to, for example, no more than $10 in any one day, $50 in any one week, and $100 in any one month. Then at least I won't have more than $100 going into thin air. I'm paying $125 for fixed bandwidth dedicated hosting, now. I can call them up and have another machine in place in an hour, and configure it in another hour. Trouble is, I have to commit a whole month for that (and I can understand that... or pay them for work actually done). Apparently the benefit of the cloud is doing this in software (VMs I assume).
It looks like they do this, but they do this BADLY. They allow firewall rules to specify which IP addresses can connect. This is bad for people on dynamic IPs. No idea, yet, if IPv6 is even supported.
What they SHOULD do is allow the user to change the SSH port number
, which is sufficient obscurity for the next few years. Hopefully, AWS will be blocking the various probers that try to dictionary attack the SSH ports, and help this situation.
Hell, even this is an oversized bloated bootloader if all you need to do is always boot ONE system and leave it running until the cleaning crew takes your power outlet. GRUB1 was horrible thought at least it was reasonably well documented, eventually. GRUB2 was worse, and depricated GRUB1 even before they had the equivalent docs out. And LILO is not even in the running. There are a couple micro boot loaders around that work on PCs, and those would be good.
Sure, there are some people around that want dual boot or more (I've built a machine with 36 OSes on it... yup, you can do more partitions in GPT... so I know what that's like). Those people might need GRUB2. But I still did the 36 OS box with Syslinux (all OSes wear Linux... no Redmond garbage here).
A shim should be a basic and simple as possible. GRUB just isn't even close.
I'm the CFH (customer from hell) that the sales people hate. I can describe the particulars and factors exactly, and then I expect them to pull a price out of their arse and be legally bound to that. Imagine you are a manager at a construction company and need a truck load of sand for a job, and the supplier tells you it costs 3.2 microdollars per grain of sand.
Reality pricing is better. Give case studies of what people actually do. Describe a web site run on a cloud, and include all the particular factors for that one, and how much it costs. Repeat for several varieties of things business, people, and spammers do on the cloud. FYI, don't plan to run a mail server there since all the spammer location aware mail servers won't take mail from a cloud based source.
What are you transmitting over these local connections? Are you backing up the contents of your cloud instances every 10 minutes? Is your cloud instance mounting your local drive for data to deliver to the web? Or is it mounting the other way around where you are using the cloud as a file server?
I haven't spent much time with AWS docs, either. That's because I can't seem to find what answers my questions. So far this seems to NOT be a true virtual machine, or real machine for that matter. Things I'd need to know is how I can add a bunch of packages (need root access) and my applications... and then CLONE it. Right now, we do this on real machines in a small farm. But it takes over 8 minutes to get a new box up and running. We're looking for something faster. But our cloning method is effectively installing the WHOLE operating system. If I take a vanilla Ubuntu install and apply all the additions and changes in place, it takes over 45 minutes to set it up. That's no good for a dynamic instance expansion.
e) I completely... 100%... recognize that the/. audience most likely prefers things like DD-WRT, Tomato, etc (though some will really like the mobile Cloud concept, I do, and I've been around the block a few times at this point). Cisco Linksys is definitely moving more towards the average consumer market instead of the tech adopter market with these products.
It should always be an owner choice (just provide a means to load a non-Cloud firmware w/o any need for a Cloud userid, for anyone with physical access to the router that can do a reset-boot). Company philosophy to the contrary means people avoiding the company for other things. Sure, Cisco most likely will not see any dings to high end core router sales. But sales of mid-level stuff that isn't even marketed to SOHO could be affected. And we may well end up prohibiting our telecommute staff from using Linksys. I've been considering the idea that the company provide routers for them, anyway. It won't be CIsco. Most likely it will be something based on DD-WRT, and maybe even a local build of it if I get the time to delve into it.
FYI, the general simple solution for a reset-boot for devices with a flash drive (USB, SD card, etc) connection is when powering up while holding the reset button, while the flash media is attached, the unwritable boot PROM will scan the media for a firmware image file in a designated directory and load it to RAM and run it. It will then default configure to default IP addresses on the LAN side. The first web page will have an option to rewrite the firmware flash storage with this firmware. This should be on all devices not intended for classified sealed operation (that's not SOHO).
f) We do still sell non-Cloud routers, like the E900, E1200 and E2500
And has the company issued a promise that these will never be borged? Keep in mind that if they ever break a promise (and I'm not sure they haven't, already) then promises from them would not longer be worth anything across the entire product line.
I've always been having troubles with Linksys. I was amazed a "reputable" company like Cisco would manage to position their low end product line at the very bottom of the low end SOHO market. But they did. And now I can't even trust them for the mid-level stuff. I'm now getting ethernet handoffs from upstream, and there are plenty of other choices like these, or just build it yourself (I have one such router working now in a data center to satisfy IPv6 needs to see a router).
Kinda like those old WinModems things that were just a cheap-ass modulator circuit driven by software that ate up your CPU time? We called those DE-celerated modems because it just didn't do the job in the modem. Are you saying NVIDIA is pulling the same shit? If so, then that's what needs to change. They need to put it all inside the video card and work from an interface that is open. Maybe they should put an X Windows or Wayland implementation directly inside the video card and just set up a bus speed network-like message path. The reference Wayland code is MIT licensed so there is no issue with them embedding a modified form of it in the video card firmware and keeping that source to themselves.
All proprietary code should be run in separate CPU and memory space apart from open source (e.g. user controlled) code in its own CPU and memory space. This is what interfaces are all about.
I still don't see how this is a "trick". Short scale fast forward just jumps frames in the buffer. Nothing tricky there. If it is a stored content, such as Youtube, you can make a new connection and specify a point in the middle. But I think what might really be going on is the added info tells the player what byte the first frame after a commercial is located at. That makes it simple to jump directly. No trick involved.
Multi-frame compression might have to jump to a group-of-pictures. No big deal there.
Then I guess the "decently employed" stipulation kicks in. I'd love to throw in the "get a job" remark here, but in this economy, it would be too insulting. Sounds like you got one, anyway, but are being shafted by the man just because he can.
Electronic voting speeds up the results. But it's only the new media that wants that.
The design I proposed was a triple path election system. There would be simple machines to vote at that produce three "results": paper, storage, and communication. But it is the paper result that counts. The stored results (on a CF card) are just for verification. The communicated results are just for the media. The paper result is actually handed to the voter. It will be printed in clear text with the names of who they voted for, and a bar code or QR code to checksum the vote. They take the paper over to the ballot box area. But first, the paper is scanned by a reader right there. Then the paper is inserted into the sealed ballot box. The scanner also stores results and transmits these results separately, which are cross checked. The official results will be the paper count. But the electronic results satisfy the media hunger for instant answers.
Slashdot Mirror
See, that's the fallacy.
Whether I would tie Lua to Apache, or Lighttpd, or nginx, or libhttp, or whatever, depends on other things. If it's a web site running other stuff, something like Apache might be the way to go. If I can run the Lua stuff on its own IP address or port, I could just run Lua the way it was designed, as an extension language for something else, like a lean HTTP daemon in C.
For now I have dedicated and colocated servers. Each "instance" is a whole machine with full root access. In the case of the colocated ones I own them, and I have VNC/IP access to the KVM on them, with a flash drive plugged in, and can recover or reload easily.
If I can't get root access, then I can't run my custom app.
Nope. Close. But 1 letter off.
But the configuration and operation of GRUB is a total bitch. I could not even find a document to describe the config file. And NO ... I do NOT configure things by running programs. I have more involved setups that just running programs cannot figure out. For example MY installer scripts need to generate the config files, NOT run some program that can't run in that minimalist installer environment.
Syslinux fits like a glove. GRUB is like trying to wrap a coat around your hand to keep it warm.
So then I do "sudo passwd root" and enter my random string a couple times?
As long as they don't have a usage cap facility that YOU can set your own limits on, in either computer based technical units of usage, or by dollars (even zero for the free tier), and in time frames other than just a whole month, then yeah ... your money could end up disappearing into thin air.
I should be able to set usage restrictions to, for example, no more than $10 in any one day, $50 in any one week, and $100 in any one month. Then at least I won't have more than $100 going into thin air. I'm paying $125 for fixed bandwidth dedicated hosting, now. I can call them up and have another machine in place in an hour, and configure it in another hour. Trouble is, I have to commit a whole month for that (and I can understand that ... or pay them for work actually done). Apparently the benefit of the cloud is doing this in software (VMs I assume).
It looks like they do this, but they do this BADLY. They allow firewall rules to specify which IP addresses can connect. This is bad for people on dynamic IPs. No idea, yet, if IPv6 is even supported.
What they SHOULD do is allow the user to change the SSH port number
, which is sufficient obscurity for the next few years. Hopefully, AWS will be blocking the various probers that try to dictionary attack the SSH ports, and help this situation.
Syslinux FTW!
Hell, even this is an oversized bloated bootloader if all you need to do is always boot ONE system and leave it running until the cleaning crew takes your power outlet. GRUB1 was horrible thought at least it was reasonably well documented, eventually. GRUB2 was worse, and depricated GRUB1 even before they had the equivalent docs out. And LILO is not even in the running. There are a couple micro boot loaders around that work on PCs, and those would be good.
Sure, there are some people around that want dual boot or more (I've built a machine with 36 OSes on it ... yup, you can do more partitions in GPT ... so I know what that's like). Those people might need GRUB2. But I still did the 36 OS box with Syslinux (all OSes wear Linux ... no Redmond garbage here).
A shim should be a basic and simple as possible. GRUB just isn't even close.
... if it has root shell access over ssh (e.g. that command line that all the New Linux geeks hate so much).
It's just a hosting provider with a few more automation scripts added on and a "gotcha" pricing model.
I'm the CFH (customer from hell) that the sales people hate. I can describe the particulars and factors exactly, and then I expect them to pull a price out of their arse and be legally bound to that. Imagine you are a manager at a construction company and need a truck load of sand for a job, and the supplier tells you it costs 3.2 microdollars per grain of sand.
Reality pricing is better. Give case studies of what people actually do. Describe a web site run on a cloud, and include all the particular factors for that one, and how much it costs. Repeat for several varieties of things business, people, and spammers do on the cloud. FYI, don't plan to run a mail server there since all the spammer location aware mail servers won't take mail from a cloud based source.
What are you transmitting over these local connections? Are you backing up the contents of your cloud instances every 10 minutes? Is your cloud instance mounting your local drive for data to deliver to the web? Or is it mounting the other way around where you are using the cloud as a file server?
WTF does "terminate-instance" really mean? Is it equivalent to "halt -p" or is it equivalent to "dd if=/dev/zero of=/dev/sda" ?
So they don't allow you to run rsync over ssh to make backups?
There are already employers wanting to hire people with 10 years experience managing 10000+ cloud instances.
I haven't spent much time with AWS docs, either. That's because I can't seem to find what answers my questions. So far this seems to NOT be a true virtual machine, or real machine for that matter. Things I'd need to know is how I can add a bunch of packages (need root access) and my applications ... and then CLONE it. Right now, we do this on real machines in a small farm. But it takes over 8 minutes to get a new box up and running. We're looking for something faster. But our cloning method is effectively installing the WHOLE operating system. If I take a vanilla Ubuntu install and apply all the additions and changes in place, it takes over 45 minutes to set it up. That's no good for a dynamic instance expansion.
You could help crack the password to my laptop.
... FTW! \o/
e) I completely ... 100% ... recognize that the /. audience most likely prefers things like DD-WRT, Tomato, etc (though some will really like the mobile Cloud concept, I do, and I've been around the block a few times at this point). Cisco Linksys is definitely moving more towards the average consumer market instead of the tech adopter market with these products.
It should always be an owner choice (just provide a means to load a non-Cloud firmware w/o any need for a Cloud userid, for anyone with physical access to the router that can do a reset-boot). Company philosophy to the contrary means people avoiding the company for other things. Sure, Cisco most likely will not see any dings to high end core router sales. But sales of mid-level stuff that isn't even marketed to SOHO could be affected. And we may well end up prohibiting our telecommute staff from using Linksys. I've been considering the idea that the company provide routers for them, anyway. It won't be CIsco. Most likely it will be something based on DD-WRT, and maybe even a local build of it if I get the time to delve into it.
FYI, the general simple solution for a reset-boot for devices with a flash drive (USB, SD card, etc) connection is when powering up while holding the reset button, while the flash media is attached, the unwritable boot PROM will scan the media for a firmware image file in a designated directory and load it to RAM and run it. It will then default configure to default IP addresses on the LAN side. The first web page will have an option to rewrite the firmware flash storage with this firmware. This should be on all devices not intended for classified sealed operation (that's not SOHO).
f) We do still sell non-Cloud routers, like the E900, E1200 and E2500
And has the company issued a promise that these will never be borged? Keep in mind that if they ever break a promise (and I'm not sure they haven't, already) then promises from them would not longer be worth anything across the entire product line.
I've always been having troubles with Linksys. I was amazed a "reputable" company like Cisco would manage to position their low end product line at the very bottom of the low end SOHO market. But they did. And now I can't even trust them for the mid-level stuff. I'm now getting ethernet handoffs from upstream, and there are plenty of other choices like these, or just build it yourself (I have one such router working now in a data center to satisfy IPv6 needs to see a router).
Kinda like those old WinModems things that were just a cheap-ass modulator circuit driven by software that ate up your CPU time? We called those DE-celerated modems because it just didn't do the job in the modem. Are you saying NVIDIA is pulling the same shit? If so, then that's what needs to change. They need to put it all inside the video card and work from an interface that is open. Maybe they should put an X Windows or Wayland implementation directly inside the video card and just set up a bus speed network-like message path. The reference Wayland code is MIT licensed so there is no issue with them embedding a modified form of it in the video card firmware and keeping that source to themselves.
All proprietary code should be run in separate CPU and memory space apart from open source (e.g. user controlled) code in its own CPU and memory space. This is what interfaces are all about.
I still don't see how this is a "trick". Short scale fast forward just jumps frames in the buffer. Nothing tricky there. If it is a stored content, such as Youtube, you can make a new connection and specify a point in the middle. But I think what might really be going on is the added info tells the player what byte the first frame after a commercial is located at. That makes it simple to jump directly. No trick involved.
Multi-frame compression might have to jump to a group-of-pictures. No big deal there.
Then I guess the "decently employed" stipulation kicks in. I'd love to throw in the "get a job" remark here, but in this economy, it would be too insulting. Sounds like you got one, anyway, but are being shafted by the man just because he can.
Electronic voting speeds up the results. But it's only the new media that wants that.
The design I proposed was a triple path election system. There would be simple machines to vote at that produce three "results": paper, storage, and communication. But it is the paper result that counts. The stored results (on a CF card) are just for verification. The communicated results are just for the media. The paper result is actually handed to the voter. It will be printed in clear text with the names of who they voted for, and a bar code or QR code to checksum the vote. They take the paper over to the ballot box area. But first, the paper is scanned by a reader right there. Then the paper is inserted into the sealed ballot box. The scanner also stores results and transmits these results separately, which are cross checked. The official results will be the paper count. But the electronic results satisfy the media hunger for instant answers.