Casey Liss

And if one fails, either their monitoring or your monitoring will have to tell somebody, hey, go replace this disk.

You have to deal with stuff like that.

Your server can break.

And you're out of the stuff.

If you don't have it backed up, too bad.

There is not much of a concept of like...

upgrading to a different server because like well you can you know you can copy the data off yourself and copy it onto a new server but you know the the concepts we have now like instances that can be resized like that didn't exist you know very little concept of like managed images and stuff like that but this is how servers were run by many people at that time at this scale

At that time, it wouldn't have made any sense for us to have our own data centers or our own racks and data centers.

We started off, the first few years of Tumblr, we were on two servers, maybe up to six or eight by the end of the second year or something like that.

I put the timeline all in there.

That's what they were.

They were dedicated servers.

And that's what everyone did before AWS EC2 and before compute instances, virtual instances, things that we use now.

And again, that world still does exist.

It's called bare metal servers.

But I think it's a lot less of the business now than it used to be.

Yeah, yeah.

The podcast 2.0 spec people defined a transcript tag, like a podcast colon transcript tag years ago.

It can point to – for an episode, it can point to a SRT or WebVTT file.

And when Apple Podcasts launched their transcripts, whatever that was, a year and a half, two years ago, they said, OK, well, we will do our own transcripts.