MySQL commands in install instructions (& elsewhere?) don't set/warn re: utf8, collation, etc about ixp-manager HOT 15 CLOSED

Here is a blog post about Forcing Utf-8 Compliance for All Connections just in case it helps (and just in case it mentions something not already known)...

from ixp-manager.

probably best on an open source project to handle it explicitly in the code rather than requiring some extra server magic to do it for you. otherwise yes, we need to use wtf-8 exclusively

from ixp-manager.

Agreed, I am new to the project and looking to deploy it in the US.

cheers
Brian

Brian Thompson
Senior Infrastructure Engineer // Senior Second Guesser

Direct: 503.943.6779
Mobile: 503.707.9018 // Twitter: iovation
www.iovation.com

On Fri, Sep 13, 2013 at 6:38 AM, Nick Hilliard [email protected]:

probably best on an open source project to handle it explicitly in the
code rather than requiring some extra server magic to do it for you.
otherwise yes, we need to use wtf-8 exclusively

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/60#issuecomment-24394856
.

from ixp-manager.

@nickhilliard said:

yes, we need to use wtf-8 exclusively

I fully endorse a switch to what-the-fuck-8 encoding for IXP Manager ;)

from ixp-manager.

@barryo :-D I am so going to call it that from now on.

@nickhilliard It is not so much a case of "extra server magic" as it is a case of saying

"MySQL, please believe me when I say I want unicode, and don't default to latin1
anyway, thinking I don't know what I'm doing"

Everything in IXP-M seems tuned to utf8 anyway, so it is not so much a "switch to utf8" as it is

"Explicitly telling the db to treat text as utf8 in order to align with existing
code, rather than carrying on blindly and *usually* getting away with it
because most ISPs' names use Latin letters anyway"

In a sane world MySQL should be defaulting to UTF-8 by now, but until such time in my opinion at least the initial commandline "create database" should ensure it. It is also about consistency because I found out the hard way that all the tools default to latin1 except for mysqldump which defaults to utf8, so by not using any commandline flags, I once ended up with double-encoded data which looked like spaghetti. I blogged a fixup tool for it a while ago here, which I hope to never need to use again...

from ixp-manager.

U NO LIKE latin1_swedish_ci??

from ixp-manager.

...extra note: when adding flags like --default-character-set=utf8 or --default-character-set=utf8mb4 they would now also need to be added to the database creation instructions in the newly created "continuous integration" pages...

from ixp-manager.

@rowanthorpe - you're PR #57 addressed the main issue - that IXP Manager connects with the database using utfmb4. I've updated the wiki to reflect this as part of the set-up instructions as well as the CI set up script.

As all scripts do or will eventually only interact with IXP Manager via the CLI / API / web interface, this means all connections will be utfmb4.

from ixp-manager.

is there a specific reason we're using utfmb4 instead of utf8?

from ixp-manager.

Read back through the thread. This link in particular:

http://mathiasbynens.be/notes/mysql-utf8mb4

from ixp-manager.

from ixp-manager.

Um, even if all of IXP-Manager's code uses unicode, the problem (which I actually opened this PR for) is that the instructions for manual setup don't specify unicode explicitly, and this means that in all but the latest versions of MySQL it will default to latin1 when created...

Also, for the patch I sent for the config file I specified 'utf8' (not 'utf8mb4') in order to err on the side of caution, and because I wasn't sure what kind of performance hit the mb4 version might cause. Perhaps you will want to profile a few things before deciding whether to set both that config file (and the manual install instructions which this PR discusses) to either 'utf8' or 'utf8mb4'. Either way, people manually setting their databases to latin1 (by omission) and then the script talking unicode to it will cause a world of hurt with anything other than ASCII.

from ixp-manager.

Oops, just noticed you said you updated the wiki @barryo - am looking at it now. Anyway, my comment about mb4-or-not still holds...

from ixp-manager.

Um, even if all of IXP-Manager's code uses unicode, the problem (which I actually opened this PR
for) is that the instructions for manual setup don't specify unicode explicitly

I've already updated the docs. If I missed something, please show me.

Also, for the patch I sent for the config file I specified 'utf8' (not 'utf8mb4') in order to err on the side
of caution, and because I wasn't sure what kind of performance hit the mb4 version might cause.

MySQL's utf8 is wrong, utf8mb4 is correct. UTF8 only uses the amount of bytes required. I'm happy with this choice.

from ixp-manager.

Ah, I just looked at the file previously changed by PR #57, and noticed that it has since been further updated to utf8mb4 (my change cautiously only set it to utf8, and I only mentioned utf8mb4 here, so I presumed it was still at that state). All good now then (I presume the mb4 hasn't caused too much of a performance hit, or it would have been evident by now).

from ixp-manager.