Comment by gk1 | Hacker News Reader

Comment by gk1 | Hacker News Reader

Here's the Twitter conversation between DHH and Yishan Wong, as of right now:

DHH: Story I heard about Reddit post-$50M investment includes everyone not in San Francisco getting 1 week to decide whether to move or GTFO. o.0

Yishan Wong: Yes, we are relocating ppl back to SF w/generous relo package & COL adjustment, +3mos severance for anyone who can’t make the move, and no, the timeline for the move is not 1 week, but through the end of the year. Decision was also independent of fundraising.

DHH: Very sad to hear. Especially for a company like Reddit that's all about bringing people together for causes regardless of location. Also, are you claiming that there never was a deadline of 1 week? That it was "through end of year" from the beginning?

Yishan: Intention is to get whole team under one roof for optimal teamwork. Our goal is to retain 100% of the team.

DHH: Hey, at least the shit sandwich comes with a smile and a thank you. Guess everything is peachy then.

Yishan: Originally we asked for decision in 2 weeks but realized almost immediately that was too short and extended the timeline to EOY.

DHH: Man, that is some cruel shit. Was that such that you could have a full headcount before the $50M check cleared?

I don't think this convo is entirely fair to Yishan. While I'm not familiar with Reddit's tech stack or management policies, I can't say they have been a shining example of a remote engineering team. For a site that is simply some text and comments, they have a staggering amount of downtime and timeouts.

Whatever they were doing before didn't seem to be working and maybe getting everyone in one place is the fix to that (not to say a remote team doesn' work, as GitHub gets along just fine).

It's not fair to him at all. DHH is espousing his opinion in his typical loud-mouthed fashion without taking the time to understand the facts on the ground. He had the same style of firebrand exchange with Travis Kalanick over the Lyft recruiting stories.

what facts "on the ground"? It's an internet link-posting company. It isn't engineering hardware. It's a website. Nothing more. This issue is pretty clear. Come bankrupt yourself trying to buy a house in SF and waste hours per week commuting or get fired. Reddit is being f'ing stupid. I hope they go out of business. Unless they cut work hours to compensate for commute times AND pay 80% more AND compensate for transportation costs, then Reddit can rot for all I care. That whole "Same office" nonsense has no quantifiable data to prove it's more effective. None. If I were forced to give up nearly 2 hours per day on commuting and the related tomfoolery, then I better be getting paid for that time. I'd rather spend it with my little kids than on a train on in a car. If the work quality doesn't meet the standard then fire the managers. "Remote" isn't the issue. The issue is the idiots in charge.

Fire the managers because they disagree with my point of view! All management decisions should be based on double-blind studies! Managing one of the most highly-trafficked sites on the Internet is easy because it isn't hardware!

I never said it was 'easy' I am making the case that on-site is irrelevant for this type of business. I realize TPD reports don't write themselves, however when was the last time you had to physically touch a coworker? Hanging out and playing ping pong together is irrelevant in terms of the work at hand. Besides Github successfully manages remotely and their application is far more technically demanding than a link posting service.

..which has nothing to do of the physical location of the workers. If there's downtime, that's because the CTO is a fucking idiot and doesn't know how to run a team or manage his stuff. Location has ZERO to do with it.

Communication within an engineering team has everything to do with it and many people consider co-location to be valuable for communication.

Do you have data on reddit's downtime to backup your claim? https://twitter.com/redditstatus shows the last downtime was on April 9th, and that was scheduled.

That must be for major site-wide downtime — anyone who frequently uses Reddit knows that a few times per week you'll see a funny little "You broke Reddit" cartoon instead of the page you were trying to load. The problem usually resolves in a matter of seconds, a minute or two at most, so it's not a big deal, and I don't believe these micro-outages are universal (i.e., I think it only happens to a subset of people viewing certain subreddits / hitting particular servers / etc.).

I don't think this necessarily means Reddit has crappy engineering. It could be they've just realized that 99.9999% success at serving pages is not a worthwhile goal. I mean, it's Reddit, it's not air traffic control — nobody's going to be hurt if they have to wait 90 seconds to see random crap on Reddit.

usually 500 errors (assuming the "You broke Reddit" is a 500) are not considered downtime. They are bugs and need to be addressed.

They are intermittent 500s, which point to performance issues.

http://www.reddit.com/r/announcements/comments/2hieim/when_t...

http://www.reddit.com/r/announcements/comments/2hyyhu/when_t...

From 5 days ago and last night, respectively. alienth disowns the threads after the downtime is complete so there's no solid record I can refer to (these are just posts I happened to be around for).

I visit reddit a lot. Besides the "You broke reddit" message mentioned by ColinCera there's a page with a blue background that says reddit is down. Pretty sure I saw it in this week if not the last 24 hours.

For a site that is simply some text and comments, they have a staggering amount of downtime and timeouts.

At reddit's scale, it doesn't matter what they're serving. The challeneges of serving that many pageviews are significant.

They might have ~175M page views per month, but for comparison HN can serve ~50M page-views per month from one server.

HN pages are much less featureful than reddit's.

I think you misread some stats. 175M is the number of unique visitors reddit gets per month, the page views is over 6 billion: https://www.reddit.com/about

On reddit you have a large portion of logged in users that all have their own front pages, and active subreddits, and active comment threads.

On HN you basically have people that have showdead on and those that don't and a few small comment threads going on at the same time.

It really is not accurate to compare the two.

Not to mention that reddit serves highly customized views to authenticated users.

While I understand that serving users at Reddit's scale is no small problem - from an outsiders pov, Reddit's engineering culture is very different from something like Pinterest (who also has the same problem wrt highly curated user feeds).

Despite this though, what I really meant that whatever Reddit is doing now doesn't seem to be working. Scaling back their remote workers might be the solution - in any case we will know sometime in 2015.

Remote working definitely isn't for everyone. If a company decides that they're better off under one roof because it works for them and allows them to be more productive/effective, so be it.

Then problem is that the increased effectiveness is usually a myth.

Why you you feel serving text is easier than serving other shapes of bits?

"Optimal teamwork" = filling your day with meetings meetings and more meetings!

Ooh, I'm snaffling that for some passive-aggressively pointed status updating tomorrow. Ta.

Depending on other aspects of team culture, being in the same office can minimize the frequency/setup/teardown of formal meetings.

> being in the same office can minimize the frequency/setup/teardown of formal meetings.

Aka "You won't get shit done because person B saves five seconds by asking you and costs you 5 minutes by them asking while you try to remember what you where doing".

Been there, done that, never again.

> costs you 5 minutes

Yeah, if it happens once. Context-switching is cumulative. 5 minutes for the first disruption, 10 minutes for the second, 30 for the next, and you just quit working after the fourth.

Not all same-place cultures encourage such interruptions. At their best, same-place cultures still have taboos and barriers against distruptive cross-talk. (And the sorts of IM/chat/conferencing tools most celebrated by distributed teams can deliver just as many derailing interruptions, if team habits encourage that.)

Co-present team members can often sense when a question is most welcome – fitting them into the day's normal interstices – or learn enough via ambient observation that many questions aren't even necessary.

I wish it was only 5 minutes. It can easily take an hour to get back into "the zone" (for lack of a better term) if you're running at peak performance when they interrupt

I hate open plan offices so much..

> being in the same office can minimize the frequency/setup/teardown of formal meetings

If it makes having meetings easier (by minimizing setup/teardown) it's more likely to cause an increase in the frequency of meetings.

New thread on DHH's full comments here -> https://news.ycombinator.com/item?id=8404006

Reply on news.ycombinator.com