Something about it really bothers me but I can't put my finger on it. Mostly due to onpopstate firing inconsistently so to speak. Maybe that's a lot older IDK, don't feel like looking it up.

I've just been skimming the thread so maybe I underestimated what he's trying to do. But in general I think his logic in why he should use rails is pretty spot on.

i agree with many of his points, but it's all outweighed for me by "use the technology you know well when on a short timeline."

Looked into Padrino a bit earlier today. It does look quite good and I will consider it, thanks for the heads up. If you read my wall of text, they basically did all of that but the big difference is they went the extra 1,000 miles to implement proper generators and most of the components are tied together in a way that's more beneficial than just having everything separate.

I like how they implemented the admin too. I like how you have to generate it when your models change and it just generates the code instead of giving you some nasty API where you try to fine tune each model in code (like django 1.0, not sure if that changed since then). That makes customizing it extremely simple and straight forward. There's no guess work either, it just generates the code that you would have written yourself in the end.

Really reminds me of one of the tips in the pragmatic programmer book. "Generate as much code as you can", and in this case it's not garbage code. It's a straight up time saver.

cool. if you do end up going that route, let me know how it goes

I have used several different CRMs. I think most of them suck ass (including the one I use now which sucks total ass) but that Salesforce.com is awesome.

For open-source CRM, are you not already familiar with Sugar?

I looked up some of open source stuff today. This new place is using salesforce. Hopefully I get to dig into it one day.

The main issue is combo products and components. Those toss me for a hard loop. Apparently this is a major upsell point ir the program gets this right. The tool I am using gets this, and inventory in general, way wrong.

I've been having some thoughts on the latest from Google's job hiring research and what it means programmers in terms of job search. I'm also trying out some the ideas I've been having around concept indexing.

When Google doesn’t use Google

I'm helping out a not-for-profit use their EMR software, and it's not uncommon that I want to call customer support for help.

The treasurer of the organization was worried that support may not be free. I had assumed it was, as their site seems to indicate so.

Wanted to ask: What's the standard for a software / SaaS provider? Is support usually free?

Can someone here explain the difference between a lexer and a parser well enough so that it is crystal clear? I mean I get the gist..lexer builds tokens/token types from input stream, parser parses said tokens. In theory it's all very clear and simple.
I mean I also get that lexers are for regular grammers and parsers for context free grammers. Also that lexers produce tokens, parsers (usually) syntax trees.

In practice however I find myself wondering if I should use a lexer or parser rule quite often. The difference doesn't seem to be clear cut. Are there any decent rules of thumb? I'm basically using "if it feels like some sort of word that always comes up like this...probably a token->lexer"

Let's take poker as an example...which of the following would you consider a token/use a lexer rule for:
card_value
card_suit
card
hand
pair

Edit: My current thinking is this...if I have a card represented as say "As2s" ... if I want to use the fact that it's a spade or an ace etc. my tokens would be "A", "s", "2", "s" or possibly "As", "2s". In the latter case card would be lexed and something like hand would be handled by the parser i.e.

hand : CARD CARD ;

In the former case...I could have

CARD : CARD_VALUE CARD_SUIT ;
or
card : CARD_VALUE CARD_SUIT ;

Dunno if I can even communicate the problem I'm facing :P

Edit2: This is ANTLR syntax i.e. lexer rules = upper case, parser rules = lower case.

100% depends on the contract/ agreement that is in place. Different providers can greatly differ between what they offer for support. It is pretty common tho for EMR SaaS providers to give free level-1 support, but I wouldn't be surprised if there is a limit on the amount of tickets you can put through, or have escalated.

As the support person you are speaking with.

Thanks, definitely will ask them. Seems to me also like it would be pretty standard to offer at least basic support for free.

Yea, typically there is some basic free support.

This is an interesting topic in general though, and one that I have spent a LOT of time thinking about. Not sure if anyone else is as interested as I am, but here are some of the things I have been pondering recently re:support:

Companies almost always view support as an expense, and thus, try and save as much money as humanely possible on it, things I have seen to save money:
1. Outsource L1 to India/ Singapore, etc.
2. Pay support people as contractors
3. Segregate Account Management and support completely, so churn is a responsibility of account management only
4. Limit the amount of tickets a company can open (weekly/monthly/yearly)

Some of the issues I always see:
1. When L1 support cannot fix an issue, they are 99% of the time absolutely horrible about communicating the escalation and it usually ends up with at least 1-2 days being in the dark about your issue.
2. L1 support is completely separate from Support Engineers/ Architects. As in, literally L1 is outsourced to a company that does not have anyone deeply knowledgable at all.

Its got me thinking of what a business model would look like that outsourced a company's total customer support and put L1-through-most senior support engineers "under one roof", to figure out a way to streamline communication between different levels.

It seems like most outsourced support is for L1 and escalations almost always end up back in the company's center of expertise, which is super inefficient. Basically I am envisioning a company that outsources the entire application support and utilized their experience across several companies to be able to minimize churn.

It would essentially help start-up and fast growing SaaS companies with combatting churn because while they may be experiencing a problem for the first time, a company who is doing L1-through-COE support for a lot of companies would be able to anticipate and respond much more quickly to help save their revenue and consult the company. This would also put less reliance on having super important individual employees for companies ("Rockstar Operations Director", etc.) and allow them to gain from the insight of other companies.

Obviously there would need to be some competitive balance, where the company wouldn't take two direct competitors without their knowledge/ agreement, but using the insight across non-competing channels seems like a good thing for everyone.

The velocity the product changes will drive the effectiveness of your plan. With a product having major changes 3 or more times a year there is little hope for those changes to be communicated in a manor sufficient to support them by an external organization. Often, new features are undocumented when a new version is put into production. Perhaps this is a start up "feature" and not the norm in more established SaaS type companies.

I'm thinking about adding this IM transcript as a comment. (I am not groucho)

(3:00:57 PM) Groucho: we should never touch this code
(3:01:08 PM) Groucho: from now on, not to touch
(3:01:26 PM) Groucho: code is ****? no, code is work. not to touch
(3:01:54 PM) Groucho: code not please aesthetically? ******* also not please aesthetically, yet it perform useful function. is no touch
(3:02:30 PM) Groucho: code have to many brackets? mother of person of suggest have too many brackets. we overlook this. never touching

Is that the result of outsourcing/rent-a-coder? :P

Strictly speaking, start-ups are probably not the right target. This would have to be more Series A/B range software companies that are bringing in a lot of revenue, but really need to get their numbers in a row in order to continue growing.

This business model basically exists already (tons of large software application companies outsource L1 support, but they still have the experts in-house, vs. at the company they are outsourcing to).

I'm thinking my option might be beneficial to mid-size companies, like 100-300 employees or so.

I wouldn't use a parser for either the card_value, card_suit, card or hand. I would tokenize the text and tag and nest the suitedness and connectedness as being meta data.

First, a classic example of tagging from language is people's name, like say for instance take the name Groucho Marx. He would be tagged as Groucho Marx (NN, PERSON). NN being a common tag for noun.

So let's use the example of As2s for poker and translate .

If you tokenized As2s in a sentence it would still be As2s. Tokenizing just separates words, paragraphs and sentences.

I think you might be think of part of speech-tagging. In this case, if we use a part of speech tagger on As2s.

We could tag it as cardcard (SuitSuit,relationship)
We would get A2 (SS, CONNECTED).

A parser would come in as you were playing say a hold'em hand and the flop came up 2h 3d 7c.

So now you need a parser to parse the relationship to the flop. So you'd parse the relationship A2(SS, CONNECTED) PAIR-> 2 3 7 (rainbow, 2-straight).

The Bird, Klein, Loper book is good introduction to the topics of tagging and parsers.

lol. this is incredible.

i want to frame the ******* line

I think the problems will be getting the experts out of the house and the friction of getting defects back into the company. Also, who pays the new support organization? Is it just the software company, their customers or the end user if they are different than the customer?

What does he mean on the last line. The others made sense and were awesome but I can't decipher the last line.

May be a poorly translated your mom joke.

i wasn't sure, but i thought it was like "code has too many brackets? your MOM has too many brackets..." obviously the non-ESL version loses all the charm of the original

fyp

Even series A/B companies probably aren't right for this. The problem like Kerowo mentioned is that it puts a pretty big impediment between your support people learning about new features and your engineers/product people learning about new bugs and common complaints.

I think most companies at this size have one general product (although maybe multi-faceted like different mobile apps) and that product needs to keep growing quickly.

As someone that's been doing support (since we don't have a support team yet the engineers just rotate) its been a great eye opener on how often fixing minor annoyances can really make people happy. It's also much easier to debate future usability changes because you're more familiar with how real life people are using the product.