Disqus - Latest Comments for melo

Re: Origin server connection security with Universal SSL

melo — Tue, 30 Sep 2014 07:56:24 -0000

Option 3, that's the one :)…. Please

Re: Announcing the Textbundle format (and Ulysses 3 giveaway)

melo — Wed, 24 Sep 2014 16:24:52 -0000

When I mentioned cross-platform I was referring at the "bundle" approach that we can do on Mac OS X, where a folder can be treated as a single file, sorry for not being clear on the initial post.

On a Mac, you can treat that folder as a single file, but only on the Mac.

Nothing more.

Glad to see the zipped spec, thanks.

Re: Announcing the Textbundle format (and Ulysses 3 giveaway)

melo — Wed, 24 Sep 2014 01:15:26 -0000

The directory approach works fine on Mac OS X because you can mark it as a Bundle and it will act as a single file.

But it is not cross platform.

Maybe wrapping it in a zip file like XLSX and JAR? This way it's always a single file, and you get compression for "free".

Yes, it gets more complicated on the implementers side, but that's what good Frameworks are for :)

Re: Tracking our SSL configuration

melo — Sun, 04 May 2014 05:15:51 -0000

Thanks, just in time for my needs

Re: Notes: Generating charset_table maps for Sphinx

melo — Sat, 22 Feb 2014 10:15:15 -0000

Maybe run with LANG=C and LC_CTYPE="" to disable the locale stuff?

Re: DDoS Prevention: Protecting The Origin

melo — Wed, 31 Jul 2013 00:53:36 -0000

One other way that can be used attackers that is easily forgotten: if your app sends email using SMTP, the IP address will be recorded in the Received header lines.

Either scrub the original Received line, or don't use SMTP for email submission.

Re: Notes: Tip: keep the Host header via nginx proxy_pass

melo — Thu, 11 Jul 2013 09:37:32 -0000

Thanks, missed that bit of documentation.

Re: Hacking Thy Fearful Symmetry

melo — Tue, 11 Sep 2012 05:58:07 -0000

Nice, very nice... :)

And welcome back!

Re: The most popular Perl web sites

melo — Sat, 08 Sep 2012 16:21:37 -0000

No report for search.cpan.org?

Re: How We Built Our 60-Node (Almost) Distributed Web Crawler

melo — Sat, 08 Sep 2012 02:06:49 -0000

They are using Spot instances, so $3 can get them more instances, it just fluctuates a bit.

Re: How We Built Our 60-Node (Almost) Distributed Web Crawler

melo — Fri, 07 Sep 2012 19:16:28 -0000

Good stuff.

Question: why do you need Gearman? Redis seems to have all the features needed to implement job distribution:

* BLPOP or BLPOPRPUSH
* PubSub

Just curious, nothing wrong with Gearman, it just looked like a simpler solution and one less piece of software to manage.

Re: Travis CI: Pull Requests Just Got Even More Awesome

melo — Thu, 06 Sep 2012 04:20:36 -0000

Ahs, great! :)

I'll ask them then.

Thanks!

Re: Travis CI: Pull Requests Just Got Even More Awesome

melo — Tue, 04 Sep 2012 20:48:21 -0000

Great job!

Some projects are small enough, or just a single person, and they don't use pull request-based workflow.

Would it be possible to annotate the commits for regular builds to? I would assume that the flag would appear in the main branch history.

Thanks!

Re: The Hitchhiker’s Guide to Riding a Mountain Lion

melo — Thu, 26 Jul 2012 20:21:26 -0000

Divvy doesn't need X11. Yes there is a error message when Divvy starts up complaining about X11, just update to the latest version, either via the App Store or from their site at http://mizage.com/divvy/.

Great post, thanks!

Re: Notes: git hosting sites

melo — Mon, 25 Jun 2012 05:54:46 -0000

Hi,

well, if you can afford it, Github has an enterprise solution: exactly the same interface - https://enterprise.github.com/

If not, you can look at gitorious. Nice interface, active development; they have paid services but you can also install a copy locally without any costs http://gitorious.org/gitori...

Re: Notes: git hosting sites

melo — Tue, 19 Jun 2012 04:58:04 -0000

The basic git installation already includes a git server that you can use over ssh.

Gitosis or gitolite should work fine on Mac OS X and provide a better management interface to multiple repositories.

Re: From TextMate to VIM for Rails Coders

melo — Mon, 04 Jun 2012 05:59:23 -0000

For project search, use ack http://betterthangrep.com/ . I'm sure you can bind it to a single key, or use a macro or function.

Re: Notes: Rasputine

melo — Mon, 21 May 2012 09:10:36 -0000

hmms.. I need more time... I don't have much at the moment...

Re: Just a Theory: Use of DBI in Sqitch

melo — Wed, 16 May 2012 05:54:44 -0000

I would assume that most of your users would be Perl users at least at first, so I would use DBI on the first version of the code.

This would allow you to focus on the core parts of your code, the schema management, and ignore the DB connectivity issues. Also, the Perl community that would be your first real users would allow you to iterate your code faster if the DB layer is well-known to many.

After the core code is stable, I would abstract the DB connectivity code, and create a IPC-based alternative, which would be the default henceforth.

Re: Notes: How to convert a UNIX timestamp to a date in Oracle

melo — Fri, 04 May 2012 04:43:44 -0000

UNIX epoch is GMT, no timezone, so the resulting date is also GMT, no timezone.

Re: Blio / Perl / App::ArchiveDevelCover 1.000

melo — Wed, 22 Feb 2012 05:48:56 -0000

Most excellent, thanks!

Re: Hacking Thy Fearful Symmetry - Embedding Perl in SQLite

melo — Tue, 21 Feb 2012 01:39:55 -0000

I'm pleasantly surprised that it took so little code to do it... :)

Really nice, gives me all sort of non-productive ideas...

Re: Notes: GitHub, CA errors and old curl's

melo — Fri, 17 Feb 2012 06:19:04 -0000

Hi,

I don't have access to a 10.7.x system, but I also get an empty line on 10.6.x and this is in fact mentioned in the article. Just keep reading.

Bye

Re: http://sergiobernardino.net/post/12514627777

melo — Tue, 08 Nov 2011 11:36:43 -0000

I will also be there, and if any programming happens it will be using Perl :)

See you there: https://codebits.eu/melo

Re: Notes: API design

melo — Tue, 08 Nov 2011 02:37:28 -0000

I failed to mention tradeoffs. When I speak of engineering the first thing that pops into my head is that engineering is all about tradeoffs.

And yes, there are situations where sensible defaults are probably good enough.

But as I get old, I see that my ability to keep all the details inside as background information is less likely to stay with me, but my pattern matching skills on what I see increase. So I don't remember as much, but I'm faster at reading and understanding more code. Hence the tradeoff for more explicit code instead of sensible (but not explicit) defaults.

As for examples, I keep getting back to one of my Perl modules: Async::Hooks. It allows you to create a list of callbacks to be called on some application event. Its a building block actually, to be used by apps that want to provide extension or notification points for external components, in a event-driven environment like POE or AnyEvent.

The main difference over other systems, is that any of the callbacks can also do some async work before calling the next element in the chain.

So the first callback, can start a async HTTP request to some external API and return. The next callback in the chain will not be called automatically by Async::Hooks. Instead, its the responsibility of the HTTP request callback to decide if they should call the next one (by actually doing it, explicitly calling the next callback in the chain) or canceling the chain (each chain can even have multiple outgoing paths, its not restricted to "next callback" and "cancel chain").

So the sensible default would be to call the next element in the chain (like most of the other observer pattern implementations on CPAN do), but I preferred to force the programmer to explicitly call the next element in the chain. If he fails to do so, the next requests won't be called (the part of dying is still not there, I have two competing solutions, haven't decided on which one will be used, waiting for more real-world usage on my part).

Thanks for your comment,