It’s Probably Why He’s Survived So Long ;)

“I’m not so weird to me.” - Haruki Murakami, The Wind-Up Bird Chronicle

“We are our own dragons as well as our own heroes, and we have to rescue ourselves from ourselves.” - Tom Robbins, Still Life With Woodpecker

“If it proves so, then loving goes by haps: Some Cupid kills with arrows, some with traps.” - William Shakespeare, Much Ado About Nothing

Submitted by @liamwon9:

TO:KY:OO / Liam Wong / liamwong.com

they said ugh i’m tired of you bitches

OFOTD #112

Giant isopods can roll themselves up into a ball to protect their soft bellies, akin to the protection mechanism of their on-land relative, the roly-poly.

Hi folks - long time tumblr user - love your work. Now that you are owned by Automattic - who have a long tradition of open source, I was wondering if you have any plans to open source the Tumblr iOS app?

Hello! Good question. We at Tumblr love open source as well! We don't know yet if we're going to open source anything like the iOS app. But if we do any new open sourcing work, we'll post about it on here, so stay tuned.

(cont.) furthermore, while I understand that tumblr's architecture makes efficient post lookup by ID impossible without a blog name or reblog key, couldn't you run a backfill process that took a very long time but repaired those broken URLs by very slowly searching for the unique post ID and replacing the URL with a uuid-based one? This may take a very long time to finish, but it would reduce the set of posts with broken trail items to ones that have been hard-deleted from the database.

This is a great follow up question! First of all: this is definitely in the realm of possibility and it’s something we’ve thought of before. At Tumblr, we’ve done some absolutely wild things at scale. When it comes to Posts on Tumblr, that’s where our scale gets really mind-blowing. 

In order to process something that could affect almost every Post record on Tumblr, we’d need to walk through billions of Post records. At our current resource capacity, this kind of operation can literally take weeks or months of real time, depending on what we need to change. When we start considering a process that would not only have to step through each individual Post record, but also reach out to other Post records for each record (sometimes thousands of other Post records if we’re talking about reblog trails), that’s when months can become a year or more of processing time. That’s how big Tumblr is.

Would it be worth doing that to fix an edge case? Maybe. That’s the tough part, and it’s something we struggle with internally. This is usually also when we start trying to come up with other, more clever ways of tackling the problem. In this case, maybe we could kick off an asynchronous task any time we encounter one of these broken trail items, so the next time it’s viewed, it might be fixed. But that won’t be perfect either, as you pointed out, because we’d need to begin tracking which records were really lost versus just changed.

Corfu - Greece (by Anna Jewels (@earthpeek)) 

https://www.instagram.com/earthpeek/