| Previous · Next | |
| User | Message |
|
frodwith
|
Date: 6/3/2009 1:09 pm · Subject: Email address validation · Rating: 0
Colin has very helpfully put in some improved email address validation recently. Seems like duplication of effort, though.
On the other hand, Email::Valid's check is correct, and Colin's will reject technically valid (though extremely uncommon) addresses. Since Email::Valid still only takes about .021 ms to do a check on my Macbook (on average), I would content that's very much fast enough, and the speed difference is between a speck of dust and a slightly larger (but actually correct) speck of dust What do you all think? Should we be using Email::Valid instead, or Colin's regex? |
| Back to Top |
Rate [ | ]
|
|
JT
|
Date: 6/3/2009 1:20 pm · Subject: Re: Email address validation · Rating: 0
Speed isn't the only consideration. You're also adding a prereq to an already large set of prereqs. So how much memory does this new prereq (and any of it's prereqs) use. |
| Back to Top |
Rate [ | ]
|
|
frodwith
|
Date: 6/3/2009 3:54 pm · Subject: Re: Email address validation · Rating: 0
After loading preload.perl in WebGUI beta, Email::Valid adds an additional 72K memory usage. |
| Back to Top |
Rate [ | ]
|
|
JT
|
Date: 6/3/2009 4:55 pm · Subject: Re: Email address validation · Rating: 0
That seems like an aweful lot of memory for something so simple, because we're not talking 72k, we're talking 72k times the number of mod perl processes you have running.On Jun 3, 2009, at 3:54 PM, <paul@plainblack.com> wrote: frodwith wrote: JT Smithph: 703-286-2525 x810fx: 312-264-5382 Create like a god. Command like a king. Work like a slave. |
| Back to Top |
Rate [ | ]
|
|
frodwith
|
Date: 6/4/2009 9:14 am · Subject: Re: Email address validation · Rating: 0
Hang on, I thought the entire reason we do the module preloading dance is to get shared versions of modules amonst modperl processes. If we load this before we fork, it's 72k period, right? |
| Back to Top |
Rate [ | ]
|
|
JT
|
Date: 6/4/2009 6:15 pm · Subject: Re: Email address validation · Rating: 0
You're right. I'm sorry. I'm so used to talking about memory being consumed by a particular method than I am entire modules that I forgot that.Regardless, I'm not sure the extra 72kb of memory gets us anything useful. On Jun 4, 2009, at 9:14 AM, <paul@plainblack.com> wrote: frodwith wrote: JT Smithph: 703-286-2525 x810fx: 312-264-5382 Create like a god. Command like a king. Work like a slave. |
| Back to Top |
Rate [ | ]
|
|
perlDreamer
|
Date: 6/4/2009 9:40 am · Subject: Re: Email address validation · Rating: 0
A couple of notes on intent, and history: My goal in putting the email regex into WebGUI::Utility was for better maintenance. For a while, we've had two email validation regular expressions:
(There may be other individual validators out there, I just haven't looked, but will keep an eye out and try to unify them all) The first has been used for over 5 years with no known complaints on email addresses. I went with the 2nd because it allows fewer illegal email addresses. |
| Back to Top |
Rate [ | ]
|
|
preaction
|
Date: 6/4/2009 2:31 pm · Subject: Re: Email address validation · Rating: 0
No known complaints I've always hated it, because I can't do something as simple as "root@localhost". Disregarding the argument over what is and is not an e-mail address, my opinion is that I do not want to keep coming back to this validation routine when there is a problem. I want one that works perfectly for all variations of e-mail addresses, or I want to send an e-mail to validate an e-mail address. Those are the only two possibilities that would ensure a proper e-mail address is entered and put the entire issue behind us. 70k is a bit, sure. Is it bigger or smaller than the XML or HTML parsers we rely on? Even HTML is easier to parse with a regular expression than RFC822 e-mail addresses (though I don't recommend regular expressions for either purpose). We can put this out of our hands and rely on the CPAN authors' expertise, or we can continue to have questions about our e-mail validation routine (and a small percentage of people getting frustrated over it).
|
| Back to Top |
Rate [ | ]
|
|
JT
|
Date: 6/4/2009 6:30 pm · Subject: Re: Email address validation · Rating: 0
I don't have a problem with this one particular module, and it's 72k. What I have a problem with is we keep adding modules for this and that and the other thing. So presuppose for a moment that we could find a module for each of the 70 form controls that come with WebGUI (right now, and you know it will grow). Let's say they average 50k a piece because email validation is probably bigger than most types of validation. Now we've just added 3.5MB to our master process size. And that's just form control validation. And you know that each of those modules grow in each process over time, because they all have data passing through them. Sure it's cleaned up and reused (if the module doesn't leak), but it's never released back to the OS until Apache is dead. What I'm getting at here is that we have to use a little be of judgement on each new module. It's not "I found a module, let's use it." It's "Does the functionality the module gives us outweigh the amount of memory the module uses?" The answer here is no. The current regex handles 99.8% of the cases. As for root@localhost, deal with it. |
| Back to Top |
Rate [ | ]
|
|
koen
|
Date: 6/7/2009 10:01 am · Subject: Re: Email address validation · Rating: 0
I don't have a problem with this one particular module Which is the subject of this discussion. What I have a problem with is we keep adding modules for this and that and the other thing. Which is another discussion all together. Do you agree that (aside from the every perl module we add needs separate judgement) using this presumably perfect perl module should have precedence over a self made regular expression? If this was the first time this functionality was needed, would you have considered the 72k increase in memory an issue? Koen de Jonge - ProcoliX |
| Back to Top |
Rate [ | ]
|
