WeenieCampbell.com
The Unwound Third => Site Announcements, Tips & Trouble Reports => Topic started by: Slack on June 25, 2019, 08:02:39 AM
-
Greetings All,
FYI, our host is upgrading our server tonight and the site will be down for less than an hour beginning at 9pm eastern time. Rivers and I have been testing the new server and it is VERY fast. You'll like it.
Cheers,
Slack
-
Here are the outage times around the world wide weenie web:
US Eastern: 9PM, 21:00
UK 2AM, 02:00
Western Europe 3AM, 03:00
NZ 1PM, 13:00
Sydney 11AM, 11:00
-
Thanks Rivers!
-
I forgot Afghanistan... :P
-
Slack and Rivers: Thanks for doing this, and of course for all your other work behind the scene and under the hood that keeps WC running so smoothly.
-
Thanks Stuart, much appreciated.
-
Yes, thanks, Rivers and Slack, for keeping the site up and running in good order, and better all the time. You're creating a community, which is pretty cool. Thanks!
-
Thank you Johnm. Nobody could predict where we'd all be when this started some 20 years ago. I'm convinced we'll be around for a long time. Unless the server migration crashes one minute from now...
-
Thank you Stuart for your participation and encouragement!
Yes, really, thank you John for your participation and for providing such high quality content. We're very lucky to have you as a resident pro.
I think it has been a fun project for us all... the genre deserves the lavishing.
-
Just to confirm, we are now fully migrated across to the new server. Any issues, comments, etc., please feel free to post them here. From where I'm sitting it's running noticeably quicker.
-
Thanks all. This is a great forum.
-
Some characters in pre-migration posts are not displaying, instead the server is replacing them with a ?
It seems to affect single quotes/apostrophes, double quotes, dashes, and maybe others.
We've reported it to our service providers and they are investigating.
Here is an example, note the ?'s embedded in the title and also substituted within the posts: https://weeniecampbell.com/yabbse/index.php?topic=11997.msg106401#msg106401
-
No ???s for viewers in Scotland. Maybe they fell into the Atlantic?
Sent from my iPad using Tapatalk
-
Interesting. So you're not seeing, for example, "Sleepy John was ? quite literally ? a legend", in David's first post in that thread, or "SOTM ? June 2019 ? Someday Baby / Worried Life Blues" in the title?
https://weeniecampbell.com/yabbse/index.php?topic=11997.msg106401#msg106401
Trick question on my part, I'm looking at the raw converted data as well as the browser output.
-
Hmmm - hadn’t looked specifically at that post. But now you? mention? it....
All recent posts seem to be immune though. Maybe it’s not catching?
Sent from my iPad using Tapatalk
-
Interesting. I'm looking at the raw converted data as well as the browser output.
Could be that the problem is with the server's software. It might be substituting the default code for ? in place of certain character codes in posts written before the upgrade. That's just a guess, but it is something the people at the server site would have to look into.
All we've been running for quite sometime now are beta versions (regardless of how they've been marketed), so it goes with the turf these days.
-
No, it's a literal '?' in the database, likely caused by codepage or CCSID differences as the data was copied between databases. We've been communicating with our host and their datacenter. There is no realtime substitution going on, the server is dishing up what's in the database.
We're working on it.
-
Hi Rivers: Yeah, it was early out here and I had a zillion things to do today, so I wasn't at my best when I wrote my post. Probably the egregious error was using "substituting" instead of its past tense. I'm sure software could be designed to convert in realtime, but it would be a clunky way to do things.
I think once the people at the site are aware of all the problems (problem characters) they can realign things and do the conversion again. With you guys running things behind the scenes, I'm certain there's a backup of the original database they can work from.
-
You were dead-on correct though; when the tense of 'substitute' is corrected from your initial present continuous to the simple past. The conversion process substituted a literal '?' for all the characters it couldn't convert.
it was a distinct possibility though that the database was correct and the server I/O was replacing them on the way to the browser. After all web servers convert a whole lot of other stuff at runtime. That was one of the first things we had to eliminate, which I did by looking at some problem posts' data records. Charlottzweb data center findings were the same.
The problem is this though. They blew the old server away without a backup very soon after cutover! I'm very surprised but it is what it is.
They do have an image of the server made immediately before migration, but to use it they will need to deploy the whole image on a new temporary server. That would be great, we would be able to figure out what database settings were different, locate all the affected posts and fix them, all with 100% precision and the lowest possible level of risk.
I have asked them to get back to us with what that deployment would take.
I don't see any other professional way to fix this problem. Piecemeal mass sql updates is so error prone it makes me feel queasy. Will keep you posted on status.
-
Hi Rivers: Thanks for the detailed reply. I haven't worked with converting databases / textbases on a server in the contemporary computer world, but I've done it with older files that require older equipment (like a 386 running DOS and its word processors such as WordPerfect 4.2; 5.1 & Word along with Windows 3.1 and Chinese Windows 3.1 and WinWord 2 and 6). People in academia often write something and then put it aside, only to find that many years later after they've upgraded several times and ditched their old equipment, they cannot read the files on their floppies and even if they can, they cannot access their files or convert them without losing formatting, Chinese characters (in Big 5; GB2312) when moving up to Unicode. Who wants to type this stuff all over again? (Providing one retained a hard copy or can remember it.) Once was one time too many, so they call on their crazy Ol' Uncle Stuie who never throws anything away and still knows a thing or two. But seriously, the point is that without the proper equipment and conversion software and filters, it's a real mess because unless things are set up correctly and done in proper sequence, text and formatting will not convert correctly.
If the server outfit saved an "image," that is, a sector-by-sector copy (using compression) of the media on which the WC database resided (usually a HDD or SSD--or part of industrial strength storage / memory), they should be able to restore it provided that the image isn't corrupted and the restoration software works correctly. Once they have the restored original, they can try converting small manageable chunks of the database that contain the problematic characters to try to identify the cause of the problem.
Anyway, thank you for all your hard work on this behind the scenes. It's time consuming and energy consuming and takes time away from listening to and playing music, something I think we'd all prefer you to be doing.
Over and Out,
Mr. Know-It-All ;)
-
Once they have the restored original, they can try converting small manageable chunks of the database that contain the problematic characters to try to identify the cause of the problem
They would be "we", as in "me", in all likelihood. :)
Great post Stuart, can't disagree with anything you've said. What people fail to appreciate, because it's usually so seamless, is that people are tapping away in multiple languages, many with squiggly accented characters, on keyboards all over the planet, using an 8-bit character set and all the memory limitations that implies. I could dive deeper but won't, because it gives me a mild headache. We are a victim of history and lack of foresight.
The best solution is to throw out character sets and standardize planet earth on a single expanded character set. 32 bits would cover it today so 64 bit would be my recommendation this year. We might be invaded by klingons in 2036 and/or have to incorporate god knows what. 64 bits should cover it, but based on the history of computing probably not. So let's go 128 bits.
The good news is we can fix it. Just need the original data.