I use cygwin on windows - it's actually a full POSIX environment with all the stuff you're used to (bash, ls, ps, grep, etc etc), and it has a package manager that will get you pretty much anything you can get on linux. It's a little bit of a pain to get set up.

The copy/paste semantic is that of unix, though, not osx, which means selecting text copies it and middle clicking pastes it. That may be configurable but I'm used to the unix way.

I don't think I would every work for someone who uses Windows, but you can try a VM, e.g. VirtualBox

Thanks Rusty.

The devs at this place use macs so it’s possible I suppose, but Seattle has a bunch of startups founded by ex MS people so it’s probably just a matter of time so may as well figure this out now.

Any reason you don't use the Ubuntu bash setup they have now?

AFAIK Cygwin has fallen out of favour now compared to the new shiny of WSL (stupid misleading name). But if you don't want to install things (both Cygwin or WSL are pretty serious) or aren't using the latest Win10, yes use PuTTY.

I still, after years and years, right click and open git bash terminals on windows. No tabs but yeah everything else works fine.. I guess.

15 years of tooling and experience I guess. Cygwin works fine, I install it once every 5 years or so, nbd.

A question for AWS gurus:

I have a cloudformation template that generates an AWS CodePipeline which builds and deploys one lambda. Each lambda is in its own repo. So each lambda/tracked branch gets its own pipeline. I can't find anyway around that, but I am at least sharing a CodeBuild project among the lambda pipelines.

The question is about the cloudformation stacks - should I create a new stack for each pipeline? Or should I just create a new change set on the same stack and execute it?

So far the only advantage I can think of to creating a new stack is if I want to go back and delete the pipelines. But I can also do that manually. Using a change set seems a little cleaner to avoid creating tons of stacks.

Any other considerations I'm missing?

WSL is basically a two step install, maybe three. Easy peasy

Anyone in IT: Never ever ever click on suspicious-looking emails. ZOMG what's wrong with you?

Corporate America: Just to confuse you, here's a horribly designed but legit email. Just click this image and login with your bank username and password on wherever you land. Hackers would never set up a dummy site that looks like BofA.

Certified by whom?

Well never mind on this. Apparently executing a change set deletes the previous resources, even if you have DeletionPolicy set to retain. I'm guessing that only kicks in on an actual stack delete, not executing a change set.

Also I didn't like that on create-change-set, aws-cli would return before the stack was ready. So I couldn't immediately run execute change-set or I'd get an error. I added a 15 second sleep and felt dirty about it.

All this is probably an indicator that AWS never intended a single CF stack to be used as a factory to generate different resources with each execution.

Either that or it's just AWS's weird tendency to force redundancy all over the place.

I should have responded earlier. CF stacks are usually tied to the logical resources they create - that's the whole point. You stand up stack1, then make changes to it using the same ID and it will basically "diff" the stacks and perform any changes to get from the old version to the new.

Regarding the sleep, we usually just wait until StackStatus = CREATE_COMPLETE. You can also detect errors this way because you'll get ROLLBACK_COMPLETE etc when things go bad.

We don't create cloudformation templates directly, we use a set of tools to do it (home grown) which is a pretty good idea, because cloudformation templates are awful. Our basic idea is that we have a base template and then a bunch of "service modification templates" each of which is meant to do one single thing with just a few knobs to turn for each. So you take your base template, add service mods + their params and it generates the full template.

As an example, we have a service mod to create an aurora database - one writer and one read slave. You pass in the username, password, database name, and EC2 instance size more or less, and it does all the stuff associated with creating the database and connecting it to the VPC and all that nonsense. If I decide I want a bigger instance I just change that and commit my changes and it'll rebuild the cloudformation template and deploy it.

We actually do have this in git but I'm a bit skeptical about how difficult it would be to use with no experience with it, and I also suspect there's quite a few assumptions built into it that are unique to our setup.

Thanks, yeah I'm kind of doing the same thing with the pipelines. I only have one CF template for all of them, and I feed it the parameters it needs. I just try to avoid creating a ton of things if I can help it.

For the lambdas themselves serverless does all the work of creating the CF templates.

As we get into more stuff I'll probably be generating CF templates. But for now everything else is one off so I'm just doing it through the console - infinitely quicker to experiment and POC.

When I tried to go the cygwin route in win I found it to be so slow it was basically unusable. Has that changed?

I currently have wsl set up on my home machine but there are some annoyances, like how all your files are marked executable for weird windows reasons.

Sadly, putty is probably best of breed for ssh-ing on windows.

I dunno, seems fine to me.

Break up the big 4!

So what is it now? Google, FB and a few other web pages?

Ugh, I hate trying to troubleshoot SaaS applications for "homework tests" when the error messages aren't exposed. "SSO Error" is not helpful or informative...

Yeah so my least favorite thing is when an API consumes another API and swallows all the errors.

Regarding the current monopolies... I feel like this time is different, but I guess we'll see. I hope I'm wrong.

I remember compuserve and AOL. I have some really funny memories of compuserve. It had a "gateway" to the internet but nothing direct. I remember you could connect to an FTP server via sending an email and then you'd get a response back that said "connected"

Then you'd send "ls /pub" and then you'd get a response back with the directory listing, and so on. You could download files, one at a time, it would email them to you. If you were used to direct FTP access it was the most laborious thing you could imagine. But if you were, like me, home for the summer with no real internet, what else were you going to do.

I used to have a giant book that was printed with thousands of public FTP sites (and archie/veronica sites, IRC channels and so forth). Without those I would have been lost because finding those on compuserve would have been nearly impossible. I wish I still had that book.

Yeah the current monopolies are doing a really good job of gobbling up, pushing out, destroying new competitors. See snapchat. They all just came out with a similar experience, diluted the waters and made Snapchat uncool.

It's gonna take another revolution like social media. Which could happen.

I remember trying to download porn on AOL and waiting for it to slowly draw on the screen. I actually kind of miss that because the anticipation was like a striptease. With broadband it was just instantly in your face.

One the early early days of AOL I remember scrolling through awesome-sounding porn images only to get a weird mime error. We called it "getting mimed". So frustrating.

Hah, mime. I made some software that downloaded stuff from usenet, and part of it was collecting, downloading, and combining mime documents. If you've never dealt with mime, each file is a collection of pieces, each piece is divided by a unique document string. In the headers at the top, you say what that document string is, just some string that is really unlikely to occur naturally in documents.

One day at a job I had someone asked me about an error message he saw. Something about "JerkengineDeluxePenis"

I was like, uh, don't worry about it, I'll figure it out/handle it. I recognized it as the document string for a mime library I'd written, and I guess I'd just copied it into our repo without thinking that I might have chosen an unsavory document string when I wrote it.

Haha I just went and looked, it's still the mime encoding string on like 8 or 9 of my old projects. I probably used that library all the time whenever I wanted to upload a file to a web server.

Lol.

I remember trying to put a document together manually with some usenet agent. So painful. What was the popular usenet agent called? I'd remember it if I hear it. Was it Fedora?

The lengths we used to have to go to get our porn.

Still beats slinking into the video store, having to ask to be let into the back, and having some creeper stand way to close to you while you try to decide between No Mans Land VII vs. Buttwoman Does Budapest.

On my way to work I drive by an adult video store that I went to a few times when I first moved to LA in 2000. Still going strong. They have some video booths in the back that I never checked out but makes me wonder if some weird **** is going on back there. Why the hell else would an adult video store still be in business in 2018?

Fedora was Redhat's linux distribution. The usenet client I used was "tin" but that was a unix thing. I have no idea what people used on windows.

It still is redhat's linux distro, right? Is something changing with the sale?