Er, no. It isn't. NoSQL solutions have exactly the same problems. Most just don't have query planners or statistics capable of understanding how to properly deal with the ever-changing data (and also frequently don't even have the indexing capabilities required to do much more than a single get or set efficiently, anyway--these problems tend to come up with larger queries).
Unfortunately, every other web designer today seems to think I don't want to read a bloody article, but rather to be engaged by an interactive article-reading application that's basically impossible to distinguish from native applications, except for those eighty quirks that are definitely going to be solved by morehacks.js and those new CSS perversions.
Web browser developers seem to cater towards those needs, which is how we ended up with browsers where I can run fifty gazillion floating-point instructions per second in JavaScript but it takes me five seconds to find a bookmark, three of which are spent hovering over the titlebar until I remember there's no menubar anymore.
I surf with JavaScript disabled by default and selectively enabled for a few frequently used sites that directly benefit from it. I rarely find sites that are unreadable without JS--certainly less often than I used to experience sites that were unreadable because of it. On the rare occasions I find a site that wont work without JS (most common symptom:completely blank page) my decision more often than not is to close the tab and move on with my life. I don't think I'm missing out on much and my computer's fans no longer scream constantly when my machine is idle with the usual dozens of open tabs.
I've run into several sites that have weird issues without Javascript, but they seem to always be things that could have been implemented with traditional markup: missing form components, misplaced images, things like that. I'd say about 50 percent of the time[1], however, media-focused sites with complicated image-viewing "galleries" or a more obscure video player, are totally useless without it.
It can be frustrating to have to go through this process of navigating to a site, realizing I've broken it, and then reloading with all the crap turned back on, but yeah, like you said, it's better than having my CPU revved up just to have those "SIGN UP FOR OUR NEWSLETTERS!" modals flying around the screen.
That's because the designers of most web pages want the features they provide. When you say "optional" you don't actually mean optional, you mean removed. Such a thing already exists, it's called Gopher[1].
Most websites today don't even gracefully degrade. One of the trendy blog/article sites that gets posted here regularly (it might be Medium) is just a column of text with lots of whitespace, but the text is loaded via AJAX, so without JS, you can't even read it.
HA! That would actually be lovely. Something easily-discovered, like a "isactualhypertextnotaturingtarpit" attribute for the <meta> tag (OK, maybe something less verbose) would probably solve half the problem.
After coding for several years, the two aren't all that different.
I'm not saying that I think that the UNIX WAY is ever going to go away...but I do think that extreme scale makes the programmatic approach to configuration make more sense. Instead of treating every system as a "system" you treat it as a simple programmable node among thousands of others. You are already starting to see Linux go this way with systemD. The stuff that the CoreOS people are doing with etcD, fleet, and flannel are really the future of *NIX.
Please cut me some slack...I'm NOT TRYING TO ARGUE ABOUT SYSTEMD. I'm just saying that it's oriented towards developers using API's. That's one of the reasons why admins who are used to the "one true way" dislike it so much. And they should have options if they don't want to use it. I'm just saying that cloud scale deployments are driving changes to infrastructure to make it more "programmable". I'm not even saying it's "right"...Its just a observation.
I think you have it backwards. Sysadmins are not the ones arguing in favor of "the one true way". That is squarely in the newer breed of developer/admin devops hybrid that the systemd camp is pandering.
For sysadmins the nix way of text in and text out allow systems to be as simple or complex as they need to be, because parts can be swapped, added or removed as needed.
The kernel don't care what your initial process is (you can for instance point the Linux kernel straight at the sh binary and be presented with a root shell the moment the kernel is done getting the hardware up and running), and the programs you want to run don't care either.
Thus you can run nix on anything from a dinky single core SoC to a warehouse sized compute cluster.
But systemd is pandering the latter while giving the former the middle finger. This by ignoring the text in text out loose bindings that has been the core of *nix.
I'm not trying to be facetious...What part of X11 matches the text in/text out -- small programs that do one thing -- mode? Even Linus says that "model" doesn't really apply any more.
Maybe systemD and it's author's are whatever...I think that the CoreOS people are demonstrating that programmatic administration is the best model for massive scale. That's my only point.
That's quite the most carefully hedged "please don't kill me" comment I've read in a long time :)
I suspect there's a scaling issue in here as well. Tools optimised for large numbers of systems are always going to look clunky and overdesigned when used on a single system.
With regard to APIs, I think this is one area where the availability of source is quite critical. "Control Panel" is clearly a tool that manipulates some API, as are all the snap-ins, but because I can't see the source
(I recently had to fix a WinCE issue by trawling the source to find the right registry key. While Windows has a power user community, WinCE really doesn't, and is subtly different in enough places to cause problems. Google sometimes makes me feel like I'm the only person using CE 7)
Wow. Before I start...that's cool that you are using CE. What are you using it for? (If you don't mind and have a minute.)
I get what you are saying. I don't think that the programmatic model of administration works for running something like PeopleSoft, right? You use large hosts that are very special and benefit from the traditional admin model. It's very much a "pet"...where the CoreOS model is more "cattle".
I'm more ambivalent about access to source as long as there is good documentation and debugging tools. I'm not going to fix a kernel bug at this point in my development. Maybe one day...
This needs some comment on proper usage (locking, isolation, ...) of these to avoid deadlocks (one thread inserts, other updates at the same time, ...). Without this this incomplete and of limited use.
I use MSSQL on a current project and really don't like it's query optimizer. It relies too much on statistics which means that slightly complex queries get random behavior really quickly, especially in the beginning as the database if growing.
IMO, completely unsuitable for web apps where DBA is not sitting around 24/7.
How would you suggest optimizing queries except on stats? Do you find instances where the stats are incorrect? If you're trying to do small things in a big table, have you used indexing to help target the small sets/fields you're looking at?
Not really. Postgres' planner is also statistics based. Yes that approach has problems, but the contrary, where you still use a plan that was appropriate for three rows with a couple million isn't good either. It's just a hard problem.
