EVE Technology Lab

I'm not sure its actually a Russian character (it doesn't render as one for me) but at one point in the description for XCAPITALSX there was a character which did not get XML escaped by CCP's API server before being transmitted.

I can't remember anything more, I'm still searching my IRC logs. I know I brought it up in #eve-dev and the guy who made the API Jackknife had also seen it.

Edit: found it and confirmed its still there.
The xCPAITALSx (corp ID 320586188) still contains the character. The character is not Russian, it is actually just a plain non-printable ASCII character that is not allowed in XML. The character is 0x0C aka \f aka form feed. The API server escapes it to the entity reference & # x C ; (spaced out because the eve-o forums own't let me type it otherwise) but even in the escaped form this is not a valid character for an XML body. I've simply patched ntt's eveapi python module to nuke this character from any downloads before it hits the actual XML parser, but it still shouldn't be there.

You can load the corp's API for yourself right here and like the python library I use firefox rightly complains about the XML being bad.



See http://www.w3.org/TR/REC-xml/#charsets for reference.




Its not even hard to get the character into your description. Here is a corporation in TEST that I have director roles in. Whenever the API refreshes the corp information it should contain the character. LOLOL Not anyone it tanked our killboard and our IT guys made me take it out.

If that was the only problem with the API

As far as I have seen, they are the only corp in an alliance in Eve that has that character. My git commit log has some unkind remarks about them.

Writing that as a CEO, not a direct API user.

Is there a supported way for me to have my corp kill log fed to both BC and eve-kill without getting one alt with director rights to create the API for the second killboard?

If Squizz, is right, and I've little doubt he is, and if I understand him correctly, that's not possible.

Is that acceptable, I don't think so, kill feed is the poster child for something a memcached deployment would improve drastically and those limitation don't shed a good light on CCP Engineering.

Squizzy is, sadly, right. And so are most other complaints here.

There are numerous eccentricities with the return XMLs of the API. The lack of typeIDs returned with ownerIDs forcing extra calls, that may or may not error depending on the call and owner type and a complete lack of KillMail caching are probably the two most problematic issues at the moment.

There are also numerous idiocies in the backend. There is no caching on erronous input so that every erronous request incurrs the full overhead of raising an exception, over and over again. Just the fact that the application keeps raising errors is a very bad decision that needs to be rethrough.

And this is not acceptable for a live service that provides our customers with useful tools we do not seem to find the resources to develop. Sadly we're now in the situation, again, where we cannot find resources to maintain the API either. There is a short meeting later today where we'll go over the situation and I will be pushing for this situation to be amended rather than swept under the rug. Hopefully something will come out of that.

And we are terribly sorry for the excessive disruption over the past few days. Sometimes you just need to rock the boat a fair bit more than usual to draw attention to the leak.

A new boat? I think you meant spaceship ;-)

Desmont,

I noticed when you added the combat log to EveMon and I remember thinking "oh yay, more 119's!", but I don't fault you for it. You're only adding a feature that your users would like to see :)

Silk daShocka,

EveMon very well could be causing some of the errors, and that would explain why I've noticed a pattern in 119's on an hourly basis: Higher during EU and US prime timezones, whereas Chinese/Russian prime timezones tend to have fewer 119's. I can only speculate that more people (e.g. English speakers) have their machines turned on with EveMon running which then forces competition for the KillLog request during those hours where 119 errors are peaking.

The biggest competition, however, isn't with EveMon. It is between zKB/Eve-Kill and Battleclinic. Users who have provided their keys to each killboard merely wants to see their kills in each website. Can't really blame the user for this as they expect the API to "Just Work (TM)."

damn right!

ok , I'm a noob , I don't understand why are they banned... I didn't change my api or done anything with it since you introduced api custimizsation ... and now aftere 2 years i'm trying to introduce it on a killboard and a I get this
"Your IP address has been temporarily blocked because it is causing too many errors. See the cacheUntil timestamp for when it will be opened again. IPs that continually cause a lot of errors in the API will be permanently banned, please take measures to minimize problematic API calls from your application."

cachedUntil is a misnomer, an euphemism at best, until CCP actually start caching (and returning cached) results.