there are two things that need to be settled here it seems:

1. how is a data search performed? it it by jumping from node to node, looking for stuff. or is it by sitting back and running the search from the confines of a single node?

2. what happens to a agent at the moment it successfully either logs in or breaks in? is it uploaded to the target node, or does it sit back in the node and just a icon show up in the target node, similar to a hackers persona?

both of these seems to have multiple interpretations flying around.

interestingly, im partial to the latter on 1. and former on 2.

this because:

1. one use a program to do the searching. this to me suggests that the program can reach out to available indexing databases and request info from there without going anywhere.

2. its the behavior agents have in earlier SR versions. if they wanted to change that now, they damn well need to state so clearly. as in, when a "free" agent requests a login to a node, in reality it requests a file transfer (or more correctly a process transfer but thats a digression), and needs to present some kind of credentials for that. said credentials will tell what level of access that transfered program will run under pr default.

thats my take on it, and thats how i will respond until im clearly disproved by some kind of official text in either book or webpage.

Agents are programs too...

But apparently my Browse programs can do things Agent software can't.

I respect the solution that agents can act on nodes they aren't loaded on. It makes sense and I don't think it's wrong (but I don't say it's right either).

As for myself, I use the other solution (the one that states that agents can only run on nodes they are loaded on) because that's how it appeared to me when I first read the rules (so it's more inuititve to me) and I'm more at ease with this interpretation as far as games mechanics are concerned.

Some of you seems to think that this vision is totally flawed. I'm not here to support it against all logic and consistency and if I ever see an arugment that can convince me, I will change my mind. But so far, I haven't found any.
I was going to explain why I wasn't convinced by the arguments that were offered, but I don't know if anybody's interested and I don't want to waste my time.

Ya know... I just treat agents as though they are a hacker too (except when the RAW explicitly state otherwise). Works well enough and makes sense for me.

I know this has been a long debate back and forth. For the most part , it's been free of the kind of venom that usually pops up by now. Let's keep it that way please and keep snarky comments out.

Not nearly as bad as Master Shake's.

Tarantula, any objection to editing in another poll option on your original post?

Create a new thread and link then.

Mechanically I look at an agent and see that when I make a check or roll I will be doing the following.

Agent Skill + Program Rating

When a Hacker does "stuff" they also do

Hacker Skill + Program Rating



When I read how Agents "Can be loaded" onto a different node, I treat this more of a "Can Also", rather as a "Can Only", where the specific descriptive text is describing an additional non standard capability of Agent software.

You'll note the rules for everything else follow this pattern where there is a general case or pattern, and then where there is something specific that supplements or contravenes an existing general rule, that specific case is written explicitly. Writing RPG rules the other way is very bloated and redundant.

For a non matrix example, you can look at the Yamaha Sakura Fubuki which has special rules when using it that are not described in the general combat and firearms section.

Actually, I was thinking about this at work today, and came upon a realization.

I think that, until we get a definite canon answer about this, new GM should consider each possibility and choose which one suits him best. Here is my take of what you get with each interpretation:

1) An agent can be loaded and running on node A but act on node B:
- The same rules apply to all Matrix entities (ie. personas)
- You can have high rating IC on low rating nodes.
- You can have as much IC on a node that you want.
- Agents are "Hackers in a Box", a rating 5 agent is just as efficient as a hacker with 5 across the board. (Except that it's not "intelligent")

2) An agent has to be loaded on a node to act on it:
- Agents are different from personas
- You can't have too much IC running on a node
- IC's rating will be limited by the node's rating

Solution 1 gives powerful attacking agents but also potentially powerful protections. It's useful to allow a character to hack without being a hacker, and is especially useful if the team doesn't have any hacker. It also uses the same rule for everyone, which might make it easier for some GM.
With solution 2, agents are more of Matrix drones rather than full-blown hacker. They can still help the hacker when going with him, but they aren't that useful on their own. On the other hand, the number and rating of ICs on a node are limited.

I tried to be as objective as I could. If you think it's not correct, missing something or showing some bias please let me know.

looks about right to me.

Agent Smith is problematic regardless of whether you use Option #1 or Option #2. It's just a problem. Heck, it was problematic in earlier versions of Shadowrun with the SAKs (it was just that noone ever bothered to buy huge amounts of memory and load multiple SAKs copies into memory because the rules were so complicated that just explaining it to the GM was probably more trouble than it was worth).

In Option #2 you still have Dr. Legion Smith and DDOSmith. In Option #1 you basically have nearly every kind of Smith problem you can imagine - but at least both sides can always play.

---

For those of you who don't know, Dr. Legion Smith is a series of arbitrarily large numbers of Agent copies who all pack Medic and repair your icon whenever they can reach it. So long as you maintain a copy of your icon in a node with Dr. Legion Smith in it, you heal all Matrix damage every single Initiative Pass. DDOSmith is a hilarious abuse of Option #2 where you are simply trying to cut off Matrix activity from a location. So your entire legion of Mr. Smiths attempts to log on to a Node with a standard account. Once they do so, they start running programs - or at least try to. Since they count as programs running on the node in question in version #2 (and arguably, so too do the programs they bring with them), they'll pop into existence and immediately drop the Response of the Node down to zero - blue screen of death. The entire system has to reboot from the memory overdraw and your Agent army can keep it suppressed by simply doing it again when it comes back up.

---

Fundamentally, the fact that "an" Agent is defined as getting its own set of die rolls to take specific Matrix actions breaks the setting.

-Frank

outfit a drone with the smarts and a medikit and you have the same issues but in real ilfe.

only that there i guess your working with physcial restrictions, while this is not so in the matrix.

and another thought, i would say that some of the agent issues depends on how you interpret the skills and programs to work. as in, when using data search and similar, is the user jumping from node to node, or sitting there hamemering searches into a number of google like services?

this is not fleshed out, so it can go either way.

and something else, some people have shows problems with what one can call a spirit storm. isnt that the magical equivalent to the agent smith problem? that one can at any moment become a virtual one man army by calling on resources one have on standby or can whistle up at a moments notice?

as in, the agent smith problem isnt isolated to the matrix. its just that it may be most severe there as there are no limitations (physical space, drain and so on) or costs involved when one can spread the load of the agents across multiple nodes (kinda like having someone else take the drain of a summoning) and can pirate the agent rather then paying for it (kinda like hijacking a truckload of drones rather then buying them).

hmm, is there a comprehension test available in SR4? i cant recall reading about one but it may be my memory failing.

as in, a test where the user rolls when giving orders, and if he rolls badly it will limit the number of dice a agent or similar can deploy when doing its own test. this because badly given orders confuse the "dog brain" of the agent.

sure, one could counter that by doing rerolls or buying hits. but when one start walking down that path one run into the classical "every offense has a counter". trying to outsmart a player by writing rules may not work in the long run...

Yes, option #2 requires a bit more fluff explanation than option #1.

Also option #2 requires fixes for the Agent Smith problem while some people can be fine with option #1 allowing both sides to use this (but some GM might want to prevent this from happening with house rules or by considering the drawbacks of an abuse of this system).
But because #2 is stating that agents don't work the same way as personas, it can lead to simple fixes.

Anyway, the entire Matrix rules need GM interpretation if not houseruling...
On one hand I hope that Unwired fixes this but on the other hand I'm afraid of some of the ways they can fix it.

Do you have a solution to the Agent Smith problem, Frank?

The solution is easy and identical to most solutions in the game for problem areas:

GM: "No."

I covered some of my answers to the Agent Smith problem earlier in this thread. I didn't really go into the reason corps don't use agent swarms because I've gone at length about that elsewhere. It's two parts fluff and one part rules implications. I can rehash it if people are interested.

On the part of hackers, there are two counts against the use of agent smith storms in attack. One is simply the resources involved if you're running them from commlinks, etc. You can get extra agents, but if you're going for the silly numbers, then that actually piles up to a hefty chunk of money to throw away on an attack. As at least some of the agents will be traced, you can't really consider it a recoverable expense. It's possible to have your agents remotely loading onto other nodes via Exploit, but there are disadvantages unique to this which is why it's important to cover any problems with having them all run from a large collection of commlinks.

Of course a corp can afford a lot of commlinks to blow on an attack, but I see this as more of a political issue than a technical one - much like lots of countries have long range missiles so they avoid launching them at each other.

So when it comes to agents loading themselves onto target nodes, we have a couple of issues. Firstly, the more agents you use, the more chances the opposition gets to spot you. If you throw six agents at a node, it is six times as likely to notice an attack as with a single agent or hacker. Because I think secrecy is very important in most hacking jobs, the agent army is actually counter-productive. If a node really sees a whole mob of coordinated assailants, it can sever connections or shut down. If you follow the sequence through, it can just get this done before the agents have a chance to do anything, even when they're hacking on the fly. A GM may allow a group of agents to take a single test using the team work rules and reduce the detection risk. I personally don't, but I know Tarantula has said that he does. I don't think this is something that can be argued as its a GM judgement call as to whether team work tests are allowed, based on his view of how the agents are actually working. I would expect that the node or any IC gets a chance to detect each one of them if they all attempt to enter the node however. In either case though, the number of extra dice on a team work test is capped as per the SR4 Errata.

Of course, forcing your target to pull up the drawbridge and sever communication with the rest of the Matrix is a victory, I'm not offering complete solutions here, but I think it will often not be the victory that you are looking for.

The Sixth World is a dangerous place. Going in with a swarm of agent programs may be effective within the narrow confines of a dungeon-crawly game session, but realistically, it creates a lot more avenues for being tracked down and identified than the softly softly approach. Being noticed can get you killed, so I think it will be avoided. Remember that you have two options - the agents are running from your commlink collection (or whatever) and that can open a lot of mundane avenues of investigation as well as Matrix ones; or else the agents are all moving about independently in which case you have to ask yourself are they stopping to erase their presence in the nodes that they've loaded themselves into, what percentage of them will not be successful in that, and which nodes are they loading themselves into? With large numbers of agents, you're going to be alerting someone (even Lonestar's Matrix division) who may begin damage control or counter action, and with reduced and safer numbers of agents being used to avoid that, you diminish the purpose of having an agent swarm in the first place,

So I don't have a magic sword with which to kill the Agent Smith Army problem, but I do have a very good collection of paper cuts and lemon juice that work quite well. The problem, as Frank has pointed out, is we don't get to give this information to everyone who picks up a copy of the SR4 rules.

-Khadim.

I agree, the Agent Smith method for attack is virtually worthless. Yes, you'll get in, but they WILL be noticed, its not cost effective, and all it does is effectively shut off that node (as it can easily cut all connections/reboot). Nothing really accomplished.

Where agent smith becomes a problem is corp defense. How much do you think the yearly salary for a security spider is? Say, 100k. A full commlink of rating 4 agent, and rating 4 analyze, armor, stealth, track, attack, blackhammer costs the company 36.9k. This assumes they won't get a bulk discount either. Assuming they do this, they can buy 3 fully loaded agents with dedicated commlinks every year. IF they can write the programs themselves, then they can get instead 15 agents per year at the cost of 6,500 per commlink, and the programs are free.

What this means, is that hacking corps is suicide (as even giving each agent only 1 shot per hacker, 10-20 agents will succeed.) I in fact lean toward the 1 test per hacker routine, because of the way you can set your analyze program to check a node for new icons/users and notify you. One test per user/icon, if you win, you see it, if you don't, you don't. Each agent just needs to have an icon in each important thread (able to have up to 8 active connections) and set their analyze programs in each one. As soon as it detects something unauthorized, they jump to that icon (free action) and begin tracking/attacking/activating alerts/telling other agents its unauthorised.

This makes hacking anyone with semi-decent resources completely unfeasible.

If you spread your agent army all over the matrix (as any ddos attack does), there will be no problems with being detected. How many ddos attackers have ever been caught in the real world?

We only need a rule saying that a system is smart enough to know that more agents over some threshold is not normal and hard blocks additional connections and\or exes from starting on a node.

This rule doesnt stop a system from sending an army after intruders though.

Or from just having an army that utterly destroys any intruders.

The easiest fix would be to say that agents require specialized hardware. If you say that agents/ic/pilots must run from network appliances with dedicated neural circuits, etc. that cost thousands of nuyen then all problems go away. No more "what node do they run from?" questions, no more "agent smith army" (for hackers or defenders). . . and no GM fiat.

or just rip out the ability to upload a agent onto the net.

if you want to use a agent, you have to run it on your "home node"...

but then you cant put a agent into a johnsons comlink, and tell it to give a shout when the johnson calls the office...

Tarantula's right when he says that the problem isn't hackers attacking with agent armies, but corps defending with them. But there's a glimmer of hope in that the players are playing the hackers and not the corps, leaving it in the GMs power to justify what corps do.

I do use IC running on remote nodes to provide security elsewhere in a Matrix system, but it's actually the minority case. The metagame reason for this is that I want to create systems that are challenging but fun for the players. The in game justifcations are either rules based (as follows) or fluff. The rules issues are as follows:

1. If you have any sort of network or subscription problem, say someone turns on a jammer in your wireless office, your security on the node the IC is supposed to be protecting vanishes with a scream of static. This could be a significant problem. And there are more sorts of problem that can occur than deliberate sabotage also. If your IC is loaded on the node it's guarding, then you can be assured that node is as safe as the IC can provide. There are no easy ways around it.

2. Assuming that the corp does have a collection of IC supporting nodes from which the agents run, it can create a point of vulnerability that can be attacked. These can be shut down or interfered with. Perhaps large quantities of IC can be subverted in one go.

On the fluff side of things, I also consider the expense. As Tarantula points out, the costs are not astronomical. However, I do find the likelihood of corps writing their own software a small one. There are issues of project management, maintenance of the software, security patches, who the manager sues when security is compromised and even simple one of trust of hiring Random Programmer instead of a reputable Megacorp. Though if you're running against Renraku, then this obviously wont be an issue so the expense argument doesn't really help you in these cases.

But broadly speaking, even greedy, save every penny corps can spend fifty - sixty thousand on security software without blinking. The problem is that you can come up with very good fluff reasons why it would be worse in reality but which aren't supported by the level of detail in the rules. For example, it's reasonable to think that software ratings would degrade with time. This would represent it getting more out of date, new exploits being found, etc. But this isn't represented in the rules. It's also reasonable to assume that a corp buying a program isn't just buying that program, but support agreements, update licences, etc. This means that the money you spend on IC and even the hardware it runs on, is money that you have to keep re-spending. It still costs Microsoft the same amount of money in sending out the install DVD whether you put it on one machine or a hundred, but they still charge you more for the licence to put it on a hundred. But again, all we have in the Gear section of the book is a purchase price.

If you expand a very small amount beyond the SR4 BBB, then it becomes easyish to justify corps not using Agent Smith Armies for defence. Certainly sufficiently for a GM wanting to preserve the balance of a role-playing game by satisfying his players' need for realism. The abuse of agents by players is handled by rules balances, so between those two, you can more or less stop the ASA popping up in your game and breaking it. It is something that needs to be looked at in Unwired for clean-up though. Even a optional rule sidebar introducing Program Rating Degredation would go some way to resolving the issue.

I was going to finish there, but it occured to me that I haven't clarified where I do use remote and roving IC. For the most part, I use it only as a set of re-inforcements when the alerts are triggered. At that point, they may be activated and sent out into the system to deal with intruders. But I use them sparingly, unless the players are trying a run against a Renraku head office.

-K.

And that fix buster removes agents from PC grasp. They already cost quite a lot, and if you make pilots require it as well, then that means drones have that hardware. I'd just take the hardware outta a few dozen microdrones, and use that to make my agent army.

Knasser, the problem goes away if you think about server rooms, and assume they're still in use in 2070. (Reasons for this are ease of maintenence, stable environment, and physical security keeps all servers out of reach, instead of needing seperate defences for each server). Put the IC in i too, and plug them directly in to all the other servers. Now, getting the server you want is just as easy as disabling the IC. And if you go the hacker way, you get owned by the dozens of IC on the system. (Even buying the programs, its almost 40k per rating 4 agent, and you could drop this alot by just taking out the hacking programs, it really only need analyze, and then have the server kick out all active connections when it sees an intruder). That drops the cost for a rating 4 agent to very cheap. 6500 for a rating 4 commlink. 20k for 2 rating 4 agents. and 800 for 2 rating 4 analyze programs. Total is 27,300 for 2. Or 13,650 for one. The security comes from terminating the connection. System does firewall +4 for active alert, + system vs expoit + hacking. Chances of the hacker winning aren't great. If it fails to kick the hacker off, the system can then just reboot.

How does making agents cost thousands (!) of nuyen do anything at all to stop corporations from using them ?

And if corporations get to use them you have successfully removed the Hacker archetype from gameplay.

oh brother, did they have to open that can of worms again...

It doesn't stop them. If they cost lots of money, it just keeps them from building super armies. Corps don't have infinite resources; they can't field armies of hackers either. It doesn't stop players if they aren't too expensive, you would have to play test the nuyen amount.

Corps don't need super armies though, just a few dozen decently rated agents is enough to lockdown a node to any hacker trying to stealth his way through, and any force large enough to take down all the agents is enough for the system to just shutdown instead of trying to fight it.

And thats exactly what they do, then the connect remotely to another node, just like a hacker can, and then pwn intruders in that node, along with the other 20 agents doing the exact same thing.